Wisdom Can't Be Copy-Pasted
Originally posted on LinkedIn on 5 May 2025.
All you can hear about these days is how tools like Claude Code and Cursor are reshaping developer workflows, and what that means for engineers, especially those just starting out. I obviously have an opinion 😀
What strikes me most is how different the learning journey looks now compared to when I began.
I started out debugging clunky enterprise Java apps and early dynamic web applications in the mid-2000s. Back then, becoming a decent developer meant reading the official documentation, experimenting, and slowly building a mental model of how systems fit together. Answers were rare. Senior devs would nudge you gently in the right direction, but you had to fight for understanding. That struggle built judgment.
Then came Stack Overflow. A game-changer. Suddenly, working code was a search away. But it also encouraged a certain cargo cult approach: copy, paste, tweak, hope for the best. Some engineers kept thinking critically. Many didn't.
Now we're seeing another shift. With #VibeCoding, an engineer can prompt an LLM to create an entire application without really understanding how it works under the surface.
That's not necessarily a problem, especially as models improve. But it changes where the real engineering happens.
I was skeptical at first. LLMs often produce clunky or outdated code. But once I began experimenting with these tools more deliberately, guiding them with architectural thinking, managing context properly, I saw glimpses of their potential. My experience helps me know what to trust and what to discard. That instinct came from years spent debugging edge cases, weighing trade-offs, and building things that had to last.
That's the part we risk skipping.
The question isn't whether AI will replace junior engineers. It's how developers at any level build the context and judgment that used to come from struggling with implementation. You don't build judgment by skimming code. You build it by being accountable for what happens when it runs in production, and when someone has to extend it years down the line.
For those just starting out, this can be disorienting. The traditional path (owning increasingly complex components until you're ready for architecture) might no longer apply. But wisdom hasn't become obsolete. If anything, it's more important than ever.
We need new ways to nurture it.
For individuals, that means staying curious about how things work, questioning AI output, and reflecting on design decisions, even when you're not writing every line yourself. For teams, it means rethinking how we mentor, review, and structure work so that judgment can actually develop.
The engineers who thrive won't be the ones who rely most heavily on AI, nor the ones who reject it outright. It'll be the ones who know when to trust it, when to guide it, and when to override it.
Wisdom can't be copy-pasted. But it can still be built.