Virtual scrolling for billions of rows

Virtual scrolling for billions of rows
21 minutes by Sylvain Lesage

A React table component can display billions of rows in a browser using five key techniques. It only loads visible data on demand, renders just the visible rows in the DOM, and uses a downscaled scrollbar to navigate tables too tall for browser limits. A dual scroll mode then handles both precise local scrolling and large jumps. Finally, vertical and horizontal scrolling are separated to allow smooth keyboard navigation to any cell.

Webinar: See a RAG-Based E-Commerce Site Built Live
sponsored by Progress Telerik

Learn how full‑stack developers boost productivity by 50% with AI agents that automate layout, styling, and component generation through RAG and LLM pipelines. See how orchestration and spec‑driven workflows keep you in control of quality and consistency. Save your seat!

React router loaders and actions as integration points
5 minutes by Sergio Xalambrí

React Router loaders and actions work best as integration points, not isolated units. Testing them alone gets messy when authentication, databases, or external APIs are involved. Instead, test the business logic they rely on, like model methods and helper functions, separately. For full end to end coverage, use a tool like Playwright to test the complete user flow.

Micro frontends: When they make sense
14 minutes by Lukas Niessen

Micro frontends split a UI into independently deployable pieces owned by separate teams, solving coordination problems rather than technical ones. Several approaches exist, from iframes and page-based routing to module federation and server-side composition, each with real tradeoffs around complexity, performance, and isolation. For most teams, a well-structured monolithic frontend with clear module boundaries beats micro frontends on simplicity. The pattern pays off mainly when many teams constantly block each other, or when migrating a legacy system piece by piece.

Building Next.js for an agentic future
4 minutes by Jiachi Liu

Making Next.js visible to AI agents took a year of experimentation. The core problem was simple: agents can't see the browser, so runtime errors and rendered components were invisible to them. This led to building an in-browser agent, then MCP integration, and finally a broader shift toward treating agents as first-class users. Now Next.js forwards browser errors to the terminal, ships a compressed docs index, and uses MCP to expose internal state so agents get the right context automatically.

Building bulletproof React components
5 minutes by Shu Ding

Components built only for simple cases break when exposed to server rendering, multiple instances, portals, or newer React features. Moving browser API calls into effects prevents server crashes. Unique IDs, context, and refs solve problems with multiple instances and portals. Using state instead of performance hints like useMemo ensures correctness when React discards cached values.

Still writing tests manually? Meticulous AI is here
sponsored by Meticulous

Most teams are forced to make the tradeoff between better coverage and more maintenance effort. But top engineering teams like Notion, Dropbox, Wealthsimple and LaunchDarkly have discovered a new testing paradigm. Built by ex-Palantir engineers, Meticulous creates and maintains E2E UI tests that covers every edge case of your web app without any developer effort - making it the only tool to improve both product quality and dev velocity.

webdev

• An in-depth guide to customising lists with CSS
• Justifying text-wrap: pretty
• Announcing Interop 2026
• Modern CSS Code Snippets
• Ultra-lightweight, zero dependency, semantic HTML, CSS, JS UI library

And the most popular article from the last issue was:

• A broken heart