Markup and styling
Semantic HTML first, always. CSS with native nesting and custom properties on new projects; Sass still earns its keep on larger, older design systems. Tailwind CSS on projects where a utility-first workflow suits the team building it.
JavaScript
// TypeScript by default on anything beyond a small script
interface Project { title: string; year: number; }
React with Next.js remains our default for anything genuinely interactive and SEO-sensitive at once; Astro for content-heavy sites with only a few interactive islands; plain, framework-free JavaScript for anything smaller than that.
Backend and hosting
Node.js and Express for custom APIs, serverless and edge functions for smaller, event-driven pieces, and a genuine preference for static-first architecture wherever the content allows it - less to host, less to secure, less to go wrong at 2am.
What hasn't changed since 2015
Semantic HTML, real accessibility testing, and writing things down for whoever inherits the project next. The tools around those principles keep changing; the principles themselves have been the same since Pranta Tech's first site.
Why we keep the list short on purpose
Every year a fresh wave of tools promises to replace half the stack. Most of them are solving a problem we don't actually have, so the real work of putting together a toolkit is less about discovery and more about disqualification - deciding what earns a place next to the things that already work. The list below survived that filter for a full year of real projects, not just a weekend prototype.
- Vite over webpack for anything greenfield - the dev-server startup time alone changes how often you're willing to restart it.
- pnpm for monorepos - the content-addressable store keeps a five-project workspace under a gigabyte instead of five.
- Playwright replacing most of our Cypress suites - one API for three browsers and a trace viewer that actually explains a flaky test.
What we deliberately didn't adopt
We tried Bun in a side project and pulled back for project work - the ecosystem compatibility gaps still bite often enough to cost more time than the speed saves. We're also holding off on full server-component adoption outside Next.js until the mental model settles across frameworks. A toolkit built from "what's new" instead of "what's proven on a deadline" is how studios end up debugging their tools instead of their product.