A Plain-HTML Developer Blog Built by Prompting an AI Coding Tool
The process shows how a non-trivial, multi-section static site can be produced entirely by prompting an AI coding tool, with the developer acting as reviewer and refiner rather than author. For anyone who finds static-site generators overkill, this is a reproducible path to a live, good-looking personal page with no framework overhead.
Starting from a blank project, a developer prompted AtomCode to generate a full personal blog with a dark theme, CSS-variable-driven light/dark mode toggle, and sections for skills, projects, and work history. The site uses only HTML, CSS, and vanilla JavaScript, with IntersectionObserver for scroll-triggered animations and localStorage to persist theme preferences. Deployment targets GitHub Pages, chosen over Vercel and Netlify for its zero-cost, zero-maintenance static hosting. The writeup includes the exact prompts used, a project file tree, and fixes for common deployment snags like Jekyll build failures and git push timeouts.
AI coding tools are shifting the role of the developer from line-by-line author to prompt engineer and QA reviewer, even for frontend work that would normally require hand-coding CSS and JS interactions.
The choice of raw HTML/CSS/JS over a static-site generator is notable: it keeps the output simple enough for an AI to generate reliably and for a human to audit quickly.
Persisting theme preference in localStorage and using IntersectionObserver for animations are small touches that give the site a polished feel without adding dependencies.