I just fumbled a casual interview with an organization a few hours ago.
It exposed a brutal truth:
Being technically competent is no longer enough by itself.
AI has changed the developer skillset. It's becoming less about how fast you can write code or how much syntax you can recall, and more about how well you can:
Understand the problem
Define clear requirements
Architect the solution
Use AI agents to implement it
Review and validate the output
I couldn't impress my interviewer because I hadn't gone a step further: building a concrete, repeatable development workflow where AI is deeply integrated into the process.
Not just using the generic skills bundled with AI IDEs, but having custom agent skills built around my own workflows and requirements.
I tried covering this by talking about my private agent-resources GitHub repo, which contains markdown files explaining how I build Next.js apps, set up services and repositories as factory functions, use Zod schemas and mappers to normalize DB and input data, and even give Cursor access to these resources as references for my projects.
But that wasn't enough.
I believe I gave them the impression that I care a bit too much about syntax, and that smells of "arcade programming" — which can imply slow delivery in this AI era.
I sat down really long just thinking about this.
Knowing LeetCode, algorithms, and syntax still has value. But they aren't the whole value proposition anymore.
The workflow is shifting from:
Requirement -> Developer writes code
to:
Problem -> Specification -> Architecture -> AI implementation -> Human review -> Testing -> Validation
The developer's job isn't disappearing.
But typing code is becoming less of the differentiator.
The real advantage now comes from knowing what to build, how to structure it, how to leverage AI effectively, and, most importantly, how to tell when the AI is wrong.
Engineering judgment is becoming the new differentiator.
