Another swing at useEffect, a conversation with the team building "Kubernetes for the front end," TanStack shipping RSC in a way I didn't expect (even though I was it live at React Paris), and GitHub finally acknowledging stacked PRs exist.
New article: You really, really, really don't need an effect
Every time I post anything about React on Reddit or LinkedIn, the comments split into two camps. Half the people yell at me for using useEffect in a snippet. The other half yell at the first half.
I got tired of it. So I wrote the article I've been circling for months. There's one question that tells you whether you need an effect, and there are seven (yes, seven) cases where you think you do but really don't.
It's got a decision tree, a long list of anti-patterns with their fixes, and a Claude Code skill you can install in two commands. There's also a section specifically for the people who've never read the React docs page this is all based on. (The docs have the answer. I just get better traction on Reddit.)
Read the full article here
Podcast: Zack Chapple and Nestor Lopez on federated frontends at 150ms
New episode of Señors @ Scale is out. Zack Chapple is the CEO and co-founder of Zephyr Cloud, the platform he calls "Kubernetes for the front end." Nestor Lopez is a Platform Engineer at Zephyr, and his journey started eight years ago with Sencha.js and iframes, long before module federation was a thing.
Zack's path started at a consulting company adding module federation support to Angular for enterprises like SAP, which exposed every pain point of scaling federated architectures. That led to Medusa, and eventually to Zephyr. Nestor came in through open source contributions to TRPC and others.
We covered a lot. Why module federation is "Docker for the front end" and Zephyr is the orchestration layer. How Zephyr hooks into any bundler and deploys to the edge in about 150 milliseconds with one line of code and no CI/CD pipeline. Their reverse tree shaking technique that recomposes federated bundles into a monolith at the edge. The time Nestor deployed 5,200+ micro frontends as a single video.
We also got into their federated MCP server for enterprise AI orchestration, a TC39 proposal to fix ESM module unloading in V8, the Chrome extension for hot-swapping MFEs in any environment, mobile support via Metro, desktop via Tauri, and what it's actually like to build a startup with four kids.
Watch the full episode on YouTube
Listen on Spotify
TanStack just shipped React Server Components
And it works completely differently from what you're used to.
In most RSC frameworks, the server owns your component tree. Components run on the server by default, you opt into interactivity with 'use client', and the server decides the final shape of everything.
TanStack Start flips that. RSCs are just streams of data that you fetch, cache, and render on your own terms from the client side.
They tested it on the TanStack website. Blog pages dropped about 153 KB gzipped from client JS. Total Blocking Time on one blog page went from 1,200ms to 260ms. A few landing pages got slightly worse — pages dominated by interactive UI don't magically get faster because you threaded a server component into the tree.
Treating RSCs as a cacheable data primitive instead of a framework paradigm feels like the right direction.
Read the full announcement
GitHub is finally working on stacked PRs
If you've ever split a feature into five dependent PRs and then burned a day rebasing the whole chain every time the first one got a review, this is for you.
GitHub just previewed gh-stack, their take on stacked PRs. It's early, explicitly a preview, but it's the first time the platform itself has acknowledged this workflow exists.
Graphite, Sapling, and Git Town users already know the value. The question is whether native support changes the default behavior of teams that currently squash everything into one mega-PR and call it a day.
Check out the preview
Community reads
The Vertical Codebase by TkDodo — Dominik's case for organizing your codebase by feature, not by file type. He walks you through an evolving folder tree instead of handing you a checklist, which is exactly how this stuff should be taught.
Read it here
How We Test TanStack AI Across 7 Providers by the TanStack team — a behind-the-scenes on how they built a test harness that runs the same suite against Anthropic, OpenAI, Google, and four more. If you're building anything provider-agnostic, the deterministic testing patterns are worth stealing.
Read it here
Things You Didn't Know About Indexes by Jon Chrt — database indexes for the person who's been writing CREATE INDEX without really knowing what happens next. Covering indexes, partial indexes, and the cases where adding an index actually makes your query slower.
Read it here
Tech Leads Are Overwhelmed. Here's How to Take Back Control by LeadDev — if you've ever ended a week of back-to-back meetings with nothing actually shipped, this one's for you. Practical advice on protecting deep work, saying no without torching relationships, and what to delegate versus what to own.
Read it here
How AI Remembers and Forgets (Part 1) by Nadia Makarevich — a deep dive into how LLM memory actually works, how context windows are managed, and why your "long chat history" feature probably doesn't do what you think. Nadia's writing is always incredibly clear, and this one is no exception.
Read it here
Workshop: From Lizard to Wizard
The waitlist is closed. The 150€ discount is gone. The workshop is back to its regular 250€ price.
There are only 3-4 spots left per cohort and that's the entire inventory. Once a date fills up, it's gone as well!
Seven modules: algorithms, system design, security, accessibility, observability, design patterns, and AI.
If you've been on the fence, this is the moment.
Grab a spot here
Conferences
JS Heroes — May 14-15, 2026
JS Heroes is back in Cluj-Napoca, Romania. Community-driven, single-track, nonprofit. Consistently one of the best lineups in Europe. If you've never been, it's worth the trip.
Use code love_for_communities for 10% off.
Get tickets here
Convex Summit — June 17-18, 2026
I'm speaking at Convex Summit 2026 at Kinépolis Ciudad de la Imagen in Madrid. Two days on how architects and tech leaders navigate complex decisions, with a lineup I'm genuinely looking forward to.
Use code CONVEX26DanNeciu for 15% off tickets.
More info and tickets
That's it for this one.
If you're enjoying these, share the subscribe link with someone who'd get something out of it, or with your team on Slack:
neciudan.dev/subscribe