Vibe coding and the human developer
AI assistants are only as good as the humans who vibe with them.
I used to play the guitar. By myself, in band formation, or with my roommate at the time who studied jazz at the conservatory in Tilburg. Over the years I lost interest, but lately I have noticed some parallels with all the vibe coding I’ve been up to.
Some days I open Claude Code, point it at our Atlas stack, and start typing without much of a concrete plan. The cursor moves faster than my thoughts. I describe what I want in plain language, let the assistant write a first draft, tweak it, Q&A it, run it, and watch it take shape. It feels less like giving orders to a machine and more like improvising with another musician. The term “vibe coding” fits because it’s not only about rigid structure, but about flow: the kind you only get when you know your instrument.
For a long time, coding has been about control. Strong typing, linters, CI pipelines, code reviews. All good things, but they can make the process mechanical. With AI tools like Claude Code or Cursor, something new is happening. Next to writing code, we’re shaping ideas together with a tool that suggests, nudges, and completes. It’s fast, sometimes messy, but feels more human.
Messy AI coding sessions are the result of a lack of rules and guardrails. In my vibe coding meta I explore how we can turn AI slop into sound, tested code.
The funny thing is, AI coding only works when you already know what you’re doing. A good developer can see when the assistant is hallucinating, when a pattern doesn’t fit the architecture, or when the generated code looks brittle. Domain knowledge and experience are what make the “vibe” productive instead of chaotic. AI can fill in the boilerplate, but it can’t tell if the idea makes sense. It doesn’t know your stack, your project goals, or the subtle culture of your team. You do. You have to explain it to the model.
At OWOW we’ve built Atlas, our internal foundation that handles the complex but critical parts: structure, reusability, DevOps, deployment, configuration. That stability makes space for exploration. When the infrastructure is solid, you can afford to experiment. I can spin up a feature in an afternoon, try a few wild approaches, and still commit something maintainable before the end of the sprint. AI becomes a partner in iteration, not a replacement for skill.
That’s why vibe coding isn’t the end of craftsmanship. It’s what happens when craftsmanship meets new tools. Instead of writing code line by line, you’re shaping a direction, keeping an eye on the rhythm and tone. The human part has leveled up.
AI assistants are only as good as the humans who vibe with them. The better you know your craft, the more meaningful (and fun) this new way of working becomes.
For me, it’s not only about speed or novelty, but about feeling the flow of building.