• Import React
  • Posts
  • 👑 How to make great React Components in 2025

👑 How to make great React Components in 2025

This week we're covering expert tips on building React components in 2025, what "use client" really means, and when to override defaults in TanStack Query. We’re also looking at how Discord scales message search, switching between LLMs in React, and some surprising insights into useEffect behavior.

Cosden Solutions Logo

Hey guys,

Welcome to another edition of Import React by Cosden Solutions!

This week we're covering expert tips on building React components in 2025, what "use client" really means, and when to override defaults in TanStack Query. We’re also looking at how Discord scales message search, switching between LLMs in React, and some surprising insights into useEffect behavior.

P.S. Want to build apps like a senior React dev? Learn the real design patterns behind scalable architecture, check out Advanced Patterns React.

Let's get into the newsletter! 🤙 

⚡️ The Latest In React 

👑 Make great React Components in 2025 with these tips
Alem Tuzlak dropped a great video breaking down 8 years of React wisdom, how to build components that are clean, composable, and built to last.

🤔 What Does "use client" Do?
React Server Components flip the script, by introducing no real API, just two powerful directives: 'use client' and 'use server'. These tell your code where it runs, letting you split logic cleanly between client and server, all inside the module system. It’s a new paradigm, not just for React, but for how we think about full-stack apps. Even if you're not using RSCs, understanding this model is super important.

🌀 When You Might Need to Override the Defaults in TanStack Query
TanStack Query ships with smart, aggressive defaultsm and most of the time, that’s all you need. It’s designed to "just work" out of the box. But sometimes, those defaults need tweaking, like when your app requires custom caching, retry behavior, or refetch intervals.

🏇 React Internals: Which useEffect runs first?
useEffect is everywhere in React, but its behavior can catch you off guard, especially in deep component trees. Ever wondered why child components render last but their effects run first? This interactive blog breaks it down, diving into React Fiber and the quirks of how React commits effects.

🧠 Selecting from Multiple Models (LLMs) with AI SDK in React.js
AI’s moving fast, and so is the demand for apps that support multiple models. In this post, Robin Wieruch shows how to build a multi-model chat UI in Next.js using the ai-sdk package, letting users switch between GPT-4 Turbo, Claude 3, and more.

📈 100s of Developers Learning to Build at Scale

A few weeks in, and the feedback on Advanced Patterns React has been incredible. I’m proud of everyone diving deep into the real design patterns behind scalable, production-grade React apps.

From advanced data fetching to routing, caching, and full-stack architecture, this course goes beyond features and teaches you how to think like a senior developer.

If you’re ready to build apps that scale confidently and cleanly, this is your next step.

🧠 AI & General Programming

📁 How Discord Indexes Trillions of Messages
Back in 2017, Discord built a message search system that could scale with billions of messages using Elasticsearch and clever sharding by guilds and DMs. It worked well, until growth pushed the system to its limits, exposing fragility during restarts, upgrades, and heavy loads. Here’s how they now handle Trillions of messages.

🏋️ Avoiding Skill Atrophy in the Age of AI
AI coding assistants are boosting productivity, but at what cost? A 2025 study from Microsoft and Carnegie Mellon warns that overreliance can dull your problem-solving edge. The more you offload, the less you flex those mental muscles, leading to skill atrophy. This blog explores avoiding skill atrophy in the age of AI.

🥱 Why performance optimization is hard work
Purplesyringa dives into the gritty reality of performance optimization, not the theory, but the grind. This post is half rant, half advice, and all-too-relatable if you've ever chased microseconds through a profiler.

🤖 Alibaba’s new Qwen3 model is impressing developers
Alibaba's Qwen team just dropped Qwen3, a powerful new lineup of eight open-source LLMs, including two cutting edge mixture of experts models. The largest, a 235B-parameter beast codenamed A22B, is already outperforming big names like DeepSeek R1 and even OpenAI’s o1 on tough benchmarks like ArenaHard.

Quick Links

7 Senior React Developer Tools (Realistic Workday)

In this older video of mine, I'll reveal the 7 essential tools that power my workflow as a senior React developer.

I'll take you through a typical workday, demonstrating how these tools boost my productivity and effectiveness.

Whether you're just starting out or looking to level up your skills, these resources will help you become a more capable and efficient React developer!

Check it out below. 👇️ 

See you soon!

Darius Cosden