axliro.dev
Back to articles
Axel Libori Roch

Axel Libori Roch

· 2 min read

My (current) Thoughts About Agentic Coding

AI has been part of our daily lives as software developers for quite some time now. It is transforming and evolving our profession in ways that would have been difficult to imagine just a few years ago. Recently, however, I finally understood the difference between vibe coding and agentic coding, and I wanted to put some of my thoughts on the subject into writing.

I've gone through different stages before reaching the point where I am today. From excitement to discouragement, including moments of genuine existential doubt about whether our profession still made sense, or whether it was worth continuing to learn when we're writing less and less code ourselves.

Until I understood it.

Until I "found my place" again.

And the most surprising part is that my place hasn't disappeared, nor has it been replaced by AI. It remains the same: understanding the problem, making decisions, defining the architecture, validating solutions, and taking responsibility for the final outcome. What has changed is not our role, but the tools we use to perform it.

Rather than writing about technical details, comparing models, or diving into complex topics in a field where I don't consider myself an expert, I want to explain what agentic coding means to me today.

For me, the goal is to have the right guardrails, processes, and context in place so that the AI agent you're using can increase your productivity by generating code that you could have written yourself. In other words: without losing control.

That's it.

I believe using AI to generate code we don't understand is incredibly tempting, but it's also a mistake. At least from the perspective of a software developer. If we don't understand what is being generated, we stop being the ones directing the process and become dependent on a black box. In that scenario, our value no longer comes from building software, but from trusting that someone—or something—else has built it correctly for us.

To me, agentic coding is not about delegating software development to AI. It's about amplifying the capabilities of a developer without giving up control, understanding, or responsibility for what is being built.