A Reconciliation of Two Terms
In public opinion, vibe coding has almost become a pejorative: not looking at the code, going purely by feel, not knowing where it broke when something goes wrong. Agentic engineering is the new darling: directing AI to work with the full apparatus of specs, tests, and review. The author's observation is that in practice the boundary between the two is not clear at all—the most rigorous agent workflow is full of "let's just try it" intuitive moments, and the most freewheeling vibe coder, once comfortable, spontaneously grows acceptance habits. They share the same core: humans retreating from writing code to defining and verifying results.
A Useful Corollary
If you accept this spectrum view, several practical questions get much clearer. A team need not choose between "ban vibe coding" and "full agentification"; the question to ask is how much verification intensity each class of task needs: a one-off script is fine sitting at the left end of the spectrum, while a core service must be pulled toward the right. The most agreed-with line in the comments: what's dangerous has never been writing code by feel, it's deciding by feel that "this doesn't need testing." Discipline should be spent on risk-grading, not on slapping labels on ways of working.
via: Hacker News