My Personal Recommended Web Frontend Tech Stack (2026 Edition)

Over the past six months, I’ve gradually settled on what feels like my current “best” tech stack.

Here’s a quick overview.

Frontend Framework

It increasingly feels like Next.js is the default choice.

Yes, it has a lot of complex features and advanced concepts, so I wouldn’t immediately recommend it to someone with very little experience.

But the way it integrates frontend and backend concerns while still keeping responsibilities clearly separated is incredibly strong from a maintainability perspective.

Unless you absolutely need to write your backend in Go or Rust, staying fully within the Node ecosystem feels almost inevitable at this point.

If you’re not yet comfortable with React, starting with Vite is probably a good idea.

But once you begin building serious projects with Next.js, you’ll likely start appreciating the developer experience sooner or later.

Styling

I’ve come to appreciate shadcn/ui quite a lot.

I still like CSS Modules.

However, for serious B2B products or web services, their advantage feels relatively small compared to what modern utility-based approaches offer.

I’m still not a huge fan of Tailwind CSS philosophically.

But shadcn’s balance of style, flexibility, and maintainability is strong enough that it outweighs that hesitation.

Compared to something like Material UI or Chakra UI, shadcn feels more customizable and less opinionated in a restrictive way.

That said, I strongly believe you should properly understand CSS before relying heavily on Tailwind or shadcn.

For smaller consumer-facing projects (toC), CSS Modules can still be a very solid choice.

Linter & Formatter

For now, ESLint + Prettier is still the safest choice.

Biome is promising and very fast.

I use it in personal projects.

However, for larger projects, the lack of fine-grained configuration can become limiting.

In teams or bigger codebases, that flexibility matters.

So I generally recommend sticking with ESLint for now.

Infrastructure

Personally, I prefer Vercel + Cloudflare.

For individual developers or small-to-medium-sized projects, this setup works extremely well.

I rarely feel limited by it.

AWS is powerful, but it can feel overly complex for many use cases.

Unless you truly need that level of control, I don’t see a strong reason to start there.

Backend & Database

Supabase feels incredibly strong right now.

The free tier is generous, the documentation is solid, and the SDK is generally pleasant to work with.

In the past, I explored AWS and Firebase, and at some point Neon or PlanetScale also felt interesting.

But once you get comfortable with Supabase, it becomes difficult to justify other options for most projects.

It’s surprisingly usable for an “all-in-one” platform.

The product design behind it is thoughtful.

Of course, there are trade-offs.

Some parts of the SDK can feel inflexible, and for projects where raw performance is absolutely critical, you may hit limitations.

But for the majority of real-world applications, Supabase is more than enough.

Desktop Apps

It’s basically Electron vs Tauri.

If you want maximum convenience and ecosystem maturity, Electron is hard to beat.

If you care deeply about bundle size and performance, Tauri is attractive.

However, it can sometimes feel like it lacks certain ergonomic conveniences.

Given that bundle size differences can easily be 5–10×, the choice isn’t trivial.

Package Manager

npm or pnpm is perfectly fine.

I don’t see a strong need to use Yarn or Bun for most cases.

pnpm is production-ready and offers excellent workspace support.

That’s a clear advantage over npm.

That said, npm has improved significantly in speed and reliability.

For standard projects, it’s completely sufficient.

Understanding npm first and then moving to pnpm if needed feels like a reasonable path.

“Vibe Coding”

At the moment, Claude Code feels indispensable.

Roughly 95% of my development workflow involves AI assistance, with the remaining parts handled manually.

I use Cursor as my editor, but VSCode would work just as well.

Testing

I didn’t go deep into testing, but for most projects, Vitest + Playwright is more than sufficient.

Overall, I think following something close to the Frontend Developer Roadmap is more than enough for most people.

I’m not entirely convinced by how strongly Astro is being recommended there, but that’s a topic for another time.