Write less code, be more responsible

The big picture

A developer writing on blog.orhun.dev argues that AI-assisted programming — vibe coding, LLMs, the whole shebang — is changing how we build software. It has been reported that tools like GitHub Copilot and Claude are increasingly part of dev workflows, and that some teams can "one-shot" a SaaS project that would have taken months a few years ago. Sounds magical. Also a bit disorienting. Not talking about AI? That’s still talking about AI. So why not talk honestly about what it does to our craft?

A messy experiment

The author recounts building a cargo-tree TUI with Codex, initially handing the model full control and then feeling "lost, confused and illiterate" — a striking emotional moment that landed like a gut punch. He switched to a commit-by-commit, quality-checked approach: use AI to solve discrete problems, then read every line, understand every semicolon. Better, he says, but also boring — you end up doing continuous code review instead of the hands-on joy of creation. The takeaway: pick your spots. Use AI for the grunt work; write the fun bits yourself. Allegedly, that balance preserves both speed and satisfaction.

Quality, transparency, responsibility

Beyond ergonomics, the post is a call to responsibility. The author urges developers to experiment without shame, to be transparent about when and how they use AI, and to keep a final, human quality pass on anything shipped. It has been reported that the influx of vibe-coded apps is making it harder to keep up with new tools, and that a growing discovery list can be overwhelming. The final line is simple and sharp: less typing doesn’t mean less care. If anything, automating more code should demand more responsibility, not less.

Sources: blog.orhun.dev, Lobsters