DEV Community: fadow

I moved from Windows to Linux — here's the full story (pain included)

fadow — Sun, 31 May 2026 16:26:56 +0000

I'm a person that use Windows since I'm still a kid

Hi, I'm a SWE that been using Windows before I even understand what "girlfriend" even meaning. From good old days when every home's pc have McAfee on their pc. But then I decided to move to Linux.

The reason I move to Linux is not because AI slop or anything, it's just a random day and I have a random thought: "Why Windows?". And then the journey begin.

At first, like all other newbies, I head to community of Linux to make a question that we all know

Hey, I come from Windows, what distro do you guys suggest me to use.

And I was kinda surprise. Not like the myth of Linux user always a not-friendly community, they actually kinda friendly. They suggest me to use distro that many people using such

Ubuntu
Mint
Debian

But then a guy slide into my DMs and say

you should try fedora, they just release KDE

And without even understand what is 'KDE'. I hop on google and search "Fedora KDE". Go to the first link and download the .iso.

After burning the .iso file into the USB and boot into the USB. Oh boy the KDE UI make me drop my jaws. Not any alike the myth where "Linux user need to type 30 commands just to install browser". Fedora KDE offer me a installer with just a single click.

And after install the fedora KDE and reboot. I got my first ever step into the Linux world.

My first ever thought is "Woah this look cool, and smooth as hell"
It can do anything Windows do but smooth like butter instead slow like a snail.

But that just a first step, now is the journey.

1. Nvidia problem

Welp nvidia decided to not open source their driver anymore. So things just tear apart for Linux user. I cannot get my screen to use my card with 100% potential. I cannot get discord to screen share. Screen sometimes stuttering. Super high latency when in X11. All just pain.

2. Fedora is not the promised land

After a few weeks of daily driving Fedora, the cracks started to show.

It works fine for basic stuff — browsing, coding, watching YouTube. Smooth, stable, no complaints. But I'm not a "basic stuff" person. I tinker. I customize. I want my desktop to feel like mine, not like someone else's default.

And Fedora pushes back when you tinker. DNF (their package manager) is slow. Finding niche packages means adding 3 different COPR repos. KDE is pretty but heavy — my laptop fan spins up just opening the settings panel. Every time I fixed one thing, two other things broke.

So I spent days scrolling Reddit, lurking in Discord servers, watching YouTube videos at 2AM. And one name kept coming up.

Arch Linux.

Everyone described it the same way: "hard to install, but after that, it's exactly what you want it to be." No bloat. No hand-holding. Just you and the system.

I was scared. I was also curious.

3. The Arch install — first attempt, first fail

I downloaded the Arch ISO. Burned it. Booted.

And then I stared at a black screen with a blinking cursor.

No installer. No "click here to continue." Just a terminal. And me.

I followed the wiki. Step by step. Partition the disk with fdisk. Format with mkfs.ext4. Mount everything. pacstrap the base system. Generate fstab. Chroot in.

Somewhere around step 14, I typed the wrong thing and the system wouldn't boot. I didn't even know what I did wrong. Just a wall of text and "kernel panic."

Second attempt. Third. Fourth. I lost count.

On the fifth try — around 4AM — it booted. A black screen with white text: archlinux login: _

I literally said "oh my god" out loud to an empty room.

4. The real pain: drivers, internet, audio, video

Booting is one thing. Making it usable is another.

Nvidia — nouveau (the open source driver) gave me 720p resolution and screen tearing. Installing the proprietary nvidia driver meant figuring out kernel modules, DKMS, and praying it doesn't break on the next pacman -Syu. It broke twice. I learned to read the Arch news before every update.

Internet — iwctl to connect to WiFi from the terminal. Not hard once you know the commands. But the first time, staring at station wlan0 scan and seeing nothing? I thought my WiFi card wasn't supported. Turns out I just forgot to bring the interface up.

Audio — PipeWire vs PulseAudio. I didn't even know this was a debate. My speakers worked, my headphone jack didn't. Two days of reading wiki pages about wireplumber config before I got both working simultaneously.

Video — Discord screen sharing on Wayland. This is still not fully solved today, but on X11 it was worse. My whole screen would freeze for 3 seconds when I clicked "share." My friends heard my voice but saw a gray box. I eventually switched to OBS virtual camera as a workaround.

5. Why I stayed

Here's the thing nobody tells you about Arch.

After you go through all that pain — the failed installs, the broken drivers, the 4AM wiki reading sessions — you come out the other side and realize you understand your computer for the first time.

On Windows, when something breaks, you restart. Maybe you Google the error code, find a forum post from 2019, run some .reg file, and pray. On Arch, when something breaks, you know why. You broke it. You fix it.

And the system you end up with? It's exactly what you want. Nothing less, nothing more.

My current setup:

Arch (btw)
RiverWM — a Wayland tiling window manager that starts in 0.3 seconds
Neovim 0.12 with native LSP
fish shell with autosuggestions
Hermes Agent to handle boilerplate work

No desktop environment. No bloat. My laptop from 2021 runs smoother than my friend's 2024 Windows machine with double the RAM.

Was it worth it?

If you just want to browse the web and edit documents — stick with Ubuntu or Mint. Seriously. Don't suffer for no reason.

But if you're a developer who wants to understand their machine, who enjoys the process of building something from nothing, who looks at Windows and thinks "why?" — Arch is waiting for you.

Just clear your weekend. And maybe the next weekend too.

This is part of my tools and setup series. I also wrote about my full dev setup here.

If you're a NestJS dev tired of setup time, I built a starter kit with auth + i18n + MongoDB + Redis pre-configured: fadow.gumroad.com

Most project templates are either too simple or absurdly over-engineered. Here's what actually works for building MVPs fast.

fadow — Fri, 29 May 2026 14:25:00 +0000

Most project templates are either too simple or absurdly over-engineered. Here's what actually works for building MVPs fast.

I'm a developer who has worked with many clients from multiple domains and countries. So I used a lot of templates to boost up my speed. And from those usage, today I'm sharing the key pain points I've discovered while using templates around the internet.

1. Over-engineering

It's wonderful for a template that already cover the future scaling if the client needed. But sometimes, it adds unnecessary complexity to the template itself. Example, if I want a micro-saas template that ready to give me to MVP ASAP. Setting up a project with k8s, microservices, RabitMQ is not only time consuming, energy drawing but also being overkill. Even though that is a good stack for when going big, but for a micro-saas with target of getting MVP ASAP to have a research about the market? It's a big down side. With me, a good template is cover scaling ability, but recognize the threshold—knowing when good enough is actually good enough to stop scaling more. A monolith project with base fundamental covering that good enough so project not getting messy to quickly, but not a gigantic walking machine that handle 100k concurrent requests is a sweet spot for me.

2. No test

Untested templates are inexcusable. What's worse for a client when they receive the thing they bought just to find it in an unusable condition? This is the absolute bare minimum of an item before reaching the client, no one is willing to buy a piece of junk after all.

3. English only

It's true that English is used widely all over the world, but there are still other languages too. Making a template for a globalized language is a good thing, but with me it's kinda not enough. Just by simply adding support to one other language at least, you have easily unlocked much more the size of the client pool. From browsing marketplace sites like Gumroads, Colorlib, Envato,... I've noticed multilingual templates consistently have higher sales than templates that only support one language, and base on that we can tell that we - human really want somethings that could help us getting things done faster but requiring less effort.

4. Too many friction

We all want things that go swiftly, no one wants to have something that go slow and wasting time to wait for it to run. And the speed I'm talking about right here is the time taken of client to use the product they bought. When a product reaches a client, they may still have to wait for hours for it to build? That's a major red flag. Instead of that, a product that starts with one command, and also boots up in seconds is a perfect thing that everyone wanted.

5. Bloated

A template with such simple built in function but taking too much space is a dead product. We have evolved for years, turning a gigantic computer that a size of a building now just fit on a desk, so why we have a template that bloated like a couple of GB just for very few functions right? Compact it, de-bloat it is a must to do for better user experiences

With me, sometimes the best template is the simplest one. No need to go crazy, no need to be huge. Just covering the base fundamental things that everyone wants to avoid is the best.

What I Built

I've created templates that address all five of these problems:

Features:

Unit & integration tests
i18n localization support
Optimized for speed & small footprint
One-command Docker setup
Full documentation & guides
Separate FE (Next.js) and BE (NestJS) templates
Modern, startup-friendly tech stack

Pricing:

BE template: $8
FE template: $13
Both FE + BE: $20

Get your template on Gumroad

Built a Vietnam phone validator API. Send any phone number format

fadow — Wed, 27 May 2026 23:29:38 +0000

I build a very fast minimal API to validate VietNam phone number that you can use instantly

Input: 0912345678, +84912345678, 84-91-2345678 — all work
Carriers: Viettel, Vinaphone, Mobifone, Vietnamobile, Gmobile

Free tier available. Live on RapidAPI:

rapidapi.com

My PC setup as a Linux user

fadow — Wed, 27 May 2026 09:53:36 +0000

Arch Linux + RiverWM + Neovim — my daily driver:

OS: Arch (btw)
WM: River (Wayland, no X11)
Editor: Neovim 0.12 (native LSP, no lspconfig.nvim)
Terminal: Alacritty + fish shell
Agent: Hermes (offloads 80% of my boilerplate work)

Why this stack:
• River starts in 0.3s. No desktop environment bloat
• Neovim with blink.cmp + treesitter — instant completions
• fish shell — autosuggestions save hours
• Hermes builds starter kits while I sleep

Results:
• NestJS Vietnam Starter shipped in 17 minutes (AI-built)
• Next.js starter in 30 minutes
• Advanced NestJS pack in 15 minutes

Your tools multiply your output.

archlinux #neovim #devsetup

How I Ship a Production NestJS Backend in Under 30 Minutes

fadow — Thu, 21 May 2026 13:04:40 +0000

Every new project used to take me 2-3 days of setup before I wrote a single line of business logic.

Auth. i18n. MongoDB. Redis. Docker. Swagger. Tests. Every. Single. Time.

Here's the stack I landed on after 6 projects:

The Setup

1. One command to run everything

# docker-compose.yml
services:
  app:
    build: .
    ports: ["3000:3000"]
  mongo:
    image: mongo:7
    ports: ["27017:27017"]
  redis:
    image: redis:7-alpine
    ports: ["6379:6379"]

docker compose up -d

3 services. Zero config drift between my machine and production.

2. Auth that doesn't leak

JWT access tokens (15min) + refresh tokens (7 days, hashed in DB). Logout invalidates both — refresh token deleted from DB, access token blacklisted in Redis. No dangling sessions.

3. i18n from day one

Vietnamese users deserve proper tone marks. Two JSON files, one auto-detect from Accept-Language, one translation function. Not a 500KB library. Just ~50 lines.

{ "email_invalid": "Email không hợp lệ" }
{ "email_invalid": "Invalid email address" }

4. Validation that actually validates

ValidationPipe({ whitelist: true, forbidNonWhitelisted: true }) — one line. No more unexpected fields in your DTOs. If your API crashes in dev, it crashes the same way in prod.

5. Tests that prove it works

36 tests. Auth flow. User CRUD. i18n responses. RBAC guards. Not "it compiled, ship it." Actually verified.

The Math

At 2h/day of freelance time, setup costs ~$25/hour:

Auth + RBAC: 2 hours = $50
i18n: 1 hour = $25
Docker setup: 30 min = $12.50
Tests: 2 hours = $50
Swagger docs: 30 min = $12.50

That's $150 in setup costs per project.

The Shortcut

I packaged this exact setup into a starter kit. Clone, .env, docker compose up -d. 30 minutes instead of 2 days.

→ https://fadow.gumroad.com/l/nestjs-starter-kit-vietnam

Bilingual Vietnamese + English. 36 passing tests. Docker-ready. $10.

What's your NestJS setup ritual? Drop it below.

7 Developer Tools I Actually Pay For (And 3 That Are Free)

fadow — Wed, 20 May 2026 02:11:17 +0000

I'm cheap. I use Arch Linux, Neovim, and open-source everything. But some tools are worth paying for. Here's my list.

Paid (worth every đồng)

1. DeepSeek API (~$10/month)

My AI coding agent runs on DeepSeek v4 Pro. Cheaper than GitHub Copilot ($10 vs $19/month). Better for Vietnamese context. I use it through Hermes Agent to build boilerplates, write tests, and debug.

DeepSeek Platform — $0.14/M input tokens

2. Railway (~$5/month)

Full-stack hosting that just works. Docker Compose one-click deploy. Built-in MongoDB + Redis. No YAML hell. No Kubernetes.

Railway — free $5 credit

3. Gumroad (free + 10% fee)

I sell my NestJS/Next.js starter kits here. Free to list. They take 10% only on sales. Built-in email marketing to past customers. PayPal + card payments handled automatically.

Gumroad — free to start

4. Namecheap (~$12/year)

Domains. .dev, .app, .com — pick whatever. I use .gumroad.com for now but will switch when there's more revenue.

Namecheap — domains from $1/year

Free (surprisingly good)

5. Hermes Agent (free, open-source)

AI agent that runs in my terminal. Builds projects, manages Obsidian, schedules cron jobs. It wrote 80% of my starter kits while I was AFK. Works with DeepSeek, OpenAI, Anthropic, local models.

Hermes Agent

6. Obsidian (free)

My second brain. PARA method. Everything I know about NestJS/Next.js/Go lives here. Free. Offline. No vendor lock-in.

Obsidian

7. Neovim (free)

0.12 "Stale Earth" with native LSP. No VS Code. No JetBrains. Instant startup. Full control.

Neovim

Not worth it (tried, cancelled)

Tool	Why I cancelled
GitHub Copilot	$19/month for worse code than DeepSeek
Notion	Slow, online-only, bloated
Vercel Pro	Hobby tier is enough for my scale
Adobe Creative Cloud	GIMP + Inkscape do the job
Grammarly Premium	Free tier catches 90%

Total monthly cost

DeepSeek API:  ~$10
Railway:        ~$5
Domain:         ~$1/month (annual)
Gumroad:        free until sale
─────────────────────
Total:         ~$16/month

$16/month to run a full dev setup with AI coding, hosting, and a storefront. Not bad.

Building a NestJS project? Save 7 hours of setup with my pre-configured starter kit.

Next.js Dark Mode with Tailwind CSS v4 in 5 Minutes

fadow — Mon, 18 May 2026 03:23:50 +0000

Dark mode isn't a "nice to have" anymore — users expect it. Here's the fastest way to add it to a Next.js 16 app with Tailwind v4 and next-themes.

1. Install

npm install next-themes

2. Wrap your app

// app/layout.tsx
import { ThemeProvider } from 'next-themes';

export default function RootLayout({ children }: { children: React.ReactNode }) {
  return (
    <html lang="en" suppressHydrationWarning>
      <body>
        <ThemeProvider attribute="class" defaultTheme="system" enableSystem>
          {children}
        </ThemeProvider>
      </body>
    </html>
  );
}

Key flags:

attribute="class" — toggles .dark class on <html>, Tailwind picks it up
defaultTheme="system" — respects OS preference on first visit
suppressHydrationWarning — prevents React hydration mismatch flash

3. Configure Tailwind v4

Tailwind v4 uses CSS-based config. Add this to globals.css:

@import "tailwindcss";

@custom-variant dark (&:where(.dark, .dark *));

That's it. No tailwind.config.js. No darkMode: 'class'. Tailwind v4 just works.

4. Build the toggle

// components/ThemeToggle.tsx
'use client';

import { useTheme } from 'next-themes';
import { useEffect, useState } from 'react';

export function ThemeToggle() {
  const { theme, setTheme } = useTheme();
  const [mounted, setMounted] = useState(false);

  useEffect(() => setMounted(true), []);

  if (!mounted) return <div className="w-9 h-9" />;

  return (
    <button
      onClick={() => setTheme(theme === 'dark' ? 'light' : 'dark')}
      className="p-2 rounded-md hover:bg-gray-100 dark:hover:bg-gray-800"
    >
      {theme === 'dark' ? '☀️' : '🌙'}
    </button>
  );
}

Why the mounted check? next-themes doesn't know the theme until after hydration (it reads from localStorage). Rendering before mount causes a flash of wrong theme.

5. Use in components

<div className="bg-white dark:bg-gray-900 text-gray-900 dark:text-gray-100">
  <h1>Hello</h1>
</div>

Every Tailwind utility works with dark: prefix. No special components needed.

Bonus: System theme + manual override

defaultTheme="system" means:

First visit → matches OS setting
User clicks toggle → saves preference to localStorage
Next visit → uses saved preference, ignores OS

This is what users expect. Don't force dark mode. Let them choose.

Full setup (dark mode + i18n + auth + dashboard): Next.js Vietnam Starter Kit — $15, ready in 5 minutes.

How to Set Up i18n in NestJS — with Vietnamese Translations

fadow — Fri, 15 May 2026 09:15:07 +0000

Most i18n tutorials stop at English and Spanish. If you're building for Vietnamese users, you need proper tone marks, natural phrasing, and a setup that doesn't fight you.

Here's how I built a NestJS i18n system that auto-detects Accept-Language and returns error messages in Vietnamese or English.

1. Install nestjs-i18n

npm install nestjs-i18n

2. Create translation files

src/i18n/
├── vi/
│   ├── validation.json
│   └── response.json
└── en/
    ├── validation.json
    └── response.json

vi/validation.json:

{
  "EMAIL_INVALID": "Email không hợp lệ",
  "PASSWORD_MIN_LENGTH": "Mật khẩu phải có ít nhất 6 ký tự",
  "USER_NOT_FOUND": "Không tìm thấy người dùng",
  "EMAIL_EXISTS": "Email đã tồn tại",
  "FORBIDDEN": "Bạn không có quyền truy cập"
}

vi/response.json:

{
  "LOGIN_SUCCESS": "Đăng nhập thành công",
  "REGISTER_SUCCESS": "Đăng ký thành công",
  "LOGOUT_SUCCESS": "Đăng xuất thành công"
}

3. Register I18nModule

// app.module.ts
import { I18nModule, HeaderResolver } from 'nestjs-i18n';
import * as path from 'path';

@Module({
  imports: [
    I18nModule.forRoot({
      fallbackLanguage: 'vi',
      loaderOptions: {
        path: path.join(__dirname, '/i18n/'),
        watch: true,
      },
      resolvers: [new HeaderResolver(['accept-language'])],
    }),
    // ... other imports
  ],
})
export class AppModule {}

4. Use in exception filters

// http-exception.filter.ts
import { I18nContext } from 'nestjs-i18n';

@Catch(HttpException)
export class HttpExceptionFilter implements ExceptionFilter {
  catch(exception: HttpException, host: ArgumentsHost) {
    const ctx = host.switchToHttp();
    const i18n = I18nContext.current();
    const response = ctx.getResponse<Response>();
    const status = exception.getStatus();
    const message = exception.message;

    // Translate if the message is a dot-notation key like "validation.EMAIL_INVALID"
    const translated = i18n?.translate(message) ?? message;

    response.status(status).json({
      statusCode: status,
      message: translated,
      timestamp: new Date().toISOString(),
    });
  }
}

5. Auto-detect language from headers

The Accept-Language header resolver does this automatically. The frontend sends Accept-Language: vi or en — NestJS picks it up. Zero extra code.

Vietnamese-specific gotchas

Tone marks: nest-i18n stores JSON as UTF-8. Vietnamese characters (ắ, ễ, ở) work natively. No special handling needed.
Polite phrasing: Vietnamese error messages should sound respectful. "Vui lòng nhập email" (Please enter email) not "Nhập email" (Enter email).
Placeholders: Use {min} and {max} — they work with nestjs-i18n's interpolation:

  "NAME_LENGTH": "Tên phải có từ {min} đến {max} ký tự"

That's it. Your NestJS API now speaks Vietnamese natively.

If you don't want to set this up from scratch, I built a NestJS Vietnam Starter Kit with i18n, JWT auth, RBAC, Docker, and 36 tests — ready to go.

How I built a NestJS + Next.js starter kit with i18n (vi/en) — and why you should too

fadow — Fri, 15 May 2026 06:12:56 +0000

After setting up the same NestJS backend boilerplate for the Nth time, I realized I was spending 1-2 days just configuring auth, i18n, Docker, Redis, Swagger — before writing a single line of business logic.

So I built two starter kits. Not vibe-coded. Clean code. Tested. Production-ready.

What's inside

🔥 NestJS Vietnam Starter ($8)

JWT auth (access + refresh tokens) + RBAC (admin/user roles)
Bilingual i18n (Vietnamese + English) — auto-detects Accept-Language
MongoDB + Redis + Bull queues
Docker Compose — one command to run everything
Auto-generated Swagger docs
Rate limiting, validation pipes, exception filters
36 tests (unit + e2e) — all green

⚡ Next.js Vietnam Starter ($13)

Next.js 16 App Router + React 19 + Tailwind v4
Dark mode + bilingual i18n
Auth pages (login/register) wired to the NestJS backend
Dashboard, profile page, PayOS payment template
Fully responsive, mobile-first

Tech stack
NestJS TypeScript MongoDB Redis Docker Next.js 16 React 19 Tailwind v4

Why I built this

Most starter kits are English-only. Vietnamese devs building for local markets need i18n from day one, with proper tone marks — not machine-translated garbage. These kits ship with natural Vietnamese translations alongside English.

Also, I got tired of Googling the same setup steps every project.

Try it

No free tier yet. But 7-day no-questions-asked refund if it doesn't save you time.

Built to use. Not to flex.