Picking a Side Project Stack Without Overthinking It

Side projects die in the setup phase more often than in the build phase. The stack should get out of your way fast—familiar tools, simple hosting, and a clear path to something you can show.

Start from what you already know

If you ship Vue at work, use Vue. If you are fastest in plain HTML and CSS, that is a valid stack. Learning three new frameworks on the same repo is a hobby tax, not a shortcut.

Familiarity compounds. You already know how to debug your usual tools, where the docs live, and which patterns fail quietly. That means more of your evening goes into the product idea instead of fighting a new bundler config. Side projects are already short on time; spending the first weekend on “hello world in a language I do not know yet” is how unfinished folders accumulate.

I still make room for learning—but I isolate it. A portfolio site is a bad place to learn Rust, GraphQL, and a new CSS methodology at once. Pick one learning goal per project, or none. Shipping with tools you know is not stagnation; it is how you build a body of finished work that later makes ambitious experiments feel grounded.

A quick check I use: can I sketch the first page and have it running locally in under an hour with this stack? If the honest answer is “after I finish the tutorial series,” the stack is too new for this project. Save that energy for a dedicated spike.

Match hosting to the project

A portfolio or blog belongs on Cloudflare Pages, Netlify, or similar—low ops, edge delivery, and room for small API routes when you need them. Reach for a heavier backend only when the idea genuinely needs auth, persistence, or real-time features you cannot fake locally.

Hosting decisions often sneak in as architecture decisions. People spin up a VPS “just in case,” then spend nights patching and renewing certificates instead of writing features. For content sites and marketing pages, static (or statically generated) hosting is almost always enough. You get HTTPS, previews on pull requests, and a deploy that is basically “push to main.”

When the idea does need a server, still prefer the lightest option that fits:

• Form submissions → a form provider or a tiny edge function
• Auth → a managed auth service instead of rolling sessions from scratch
• Database → a hosted DB with a simple SDK, not a self-managed cluster

I ask one question before adding infrastructure: what breaks if this service is down for an hour? If the answer is “nothing critical,” keep it static. If the answer is “users cannot complete the core action,” then earn the ops cost.

For this site, Nuxt plus Cloudflare Pages is enough: markdown content, prerendered routes, and occasional server routes without running a long-lived Node process myself. That match of project shape to hosting shape matters more than chasing the newest platform blog post.

Boring defaults win

For most personal sites, a short list of boring defaults beats a custom architecture diagram:

• A static site generator or SPA you are comfortable with
• Markdown or JSON for content
• Git for version control
• One deploy command (or one CI job that runs that command)

That is enough to publish for years. Content in files means reviews happen in pull requests. Git means you can revert a bad post. One deploy command means you are not the only person who knows the release ritual—future you counts as “people.”

I also keep dependencies thin on purpose. Every package is something that can break on upgrade day. Prefer the platform’s built-ins and a few libraries you already trust. A side project does not need a design system monorepo, a state-management framework, and three CSS abstraction layers before the first page ships.

Example of a stack that stays boring and shippable:

Nuxt (or Vite + Vue)
Markdown in /content
npm run build → static output
Cloudflare Pages on push to main

If you need a CMS later, you can add one. Starting with markdown does not lock you out of that path; it just lets you publish before you invent editorial workflow.

When to experiment

Save new tools for a throwaway branch or a tiny spike repo. Prove the idea in a weekend with familiar tools first. Swap pieces later if a real constraint shows up—not because a blog post said you should.

Experimentation is healthy when it has a boundary. Timebox a spike: “two evenings to see if this library solves X.” Write down the success criteria before you start. If the spike fails, delete the branch without guilt. If it succeeds, migrate deliberately—one piece at a time—rather than rewriting the whole app mid-feature.

What counts as a real constraint worth swapping for:

• Build times that make iteration painful
• A missing capability you cannot fake (offline sync, realtime collaboration)
• A security or compliance need you cannot meet with the current host

What does not count: boredom with your current stack, fear of missing a trend, or the urge to rebuild because the README looks too simple. Simple READMEs are a feature.

When you do adopt something new, keep the public surface of the project stable. Users do not care that you switched CSS tools; they care that links still work. Migrate behind the same routes and content shapes so the experiment does not become an accidental redesign.

Wrap-up

Pick familiar tools, host simply, and ship something small. You can always refactor once the project has a reason to exist beyond the README—and by then, the constraints that matter will be obvious instead of imaginary.