Engineering
Delete the Framework From Your Blog
A blog is the one project where the runtime is pure overhead. Here is the argument, and the build that settles it.
Every blog framework starts by making a promise: give me a folder of Markdown and I will give you a site. The promise is real. What it costs you arrives later, in the bundle.
For a site whose entire product is readable text, the runtime is the largest single cost and it buys almost nothing.
What the runtime actually costs
Not the kilobytes, exactly — those are measurable and cheap. The cost is what it makes impossible:
- You cannot see the HTML a crawler sees without running a build.
- You cannot serve the site from anywhere but a runtime, so the host is a decision.
- A page that errors at runtime has no content at all, because the content lived in JavaScript.
- Every reader downloads code for pages they may never open.
What static gets you
A build step that renders every route to a file. The crawler gets the same bytes you do. The host is a directory on a disk. A broken deploy fails loudly at build time rather than quietly in production.
npm run build
# 1. vite build -> hashed JS + CSS in dist/assets
# 2. vite build --ssr -> a renderer in .ssr/
# 3. node scripts/prerender -> one HTML file per routeThree commands and the third one is a plain Node script using `react-dom/server`, which was already installed. No plugin, no browser download, no headless Chrome in the install.
Where this stops working
The moment a page genuinely differs per visitor — a logged-in dashboard, a cart, a comment form that has to post — a runtime is not optional, and pretending otherwise is how static sites get an awkward bolt-on API within a year.
A blog is the clean case. That is why it should be the easy one.