The LLM spectrum and responsible LLM use
Summary
This article maps a spectrum of LLM-assisted development from fully manual to 'vibe coding' and evaluates middle-ground modes (Tab autocomplete, localized prompts, Cmd+K prompts). It cautions that while some automation can speed up boilerplate tasks, heavy reliance on agents risks maintainability and the developer's mental model; the author advocates hands-on coding and selective, responsible use.