Yadiel

Today was my last day at Connect.

Super excited for what’s next!!

fuck! I'll be out of town 😢

Something analogous to TypeScript's strict flag, where developers can opt-in, without imposing it on the entire community by default, to a build that only succeeds if all the Rules of React/Hooks are satisfied.

That's a great point, and the lack of precedent is definitely a huge hurdle. It does make me wonder if a "strict mode" for the compiler could be a viable middle ground.

Hot take: Now that we have the React Compiler, hook rules should be enforced at the compiler level.

I've seen too many codebases just ignore them. I've seen hooks used conditionally, missing deps in arrays, or no eslint-plugin-react-hooks at all. This stuff should be unskippable

Too good! Honestly, that video might've been the final nudge to buy his design tokens course haha @bradfrost.com

I'd say that at least for folks building their DS with Tailwind, these type of components have become kinda unnecessary, maybe unless it's a more complex layout that it's used over and over

At my org we're doing this and it has worked well

what about a kickstarter/indiegogo campaign?

3Blue1Brown visual explanations are beautiful

Tried to visit the Next.js docs, typed nextjs.dev by mistake… and it redirects to the Remix site 😂 whoever set that up is a total troll

few weeks ago I was in the middle of a technical interview, when I opened the dev tools, it crashed. Back to Chrome

yes, I implemented it this way, and it just works!

started from the bottom (useEffect) now we're here (RSC)

pre-commit hooks are mental illness

I was in the middle of a technical interview when Arc started freezing 💀

Back to Chrome

What are RSC's tradeoffs and pain points?

then, those comments are usually followed by "a SpA iS BeTtEr"

dreaaam of enshittificatiooon🎶

Great article! I'll def try this strategy.

The only issue I see is if a project uses a dep. for multiple things, ideally I would like it to span multiple categories.

Small side note: package.json supports comments only in Bun, and Deno supports jsonc for their config file

I have given it more thought, I was wrong because we can technically have RSC in an MPA framework; it would be just like the initial page load in a Next app for example, right?

What about:

Astro is RSC with MPA

While not entirely sure if it's a RSC-exclusive feature, with RSCs you can have out-of-order streaming which allows you to send everything that’s fast or done to the browser, and slow components will be streamed in when they’re done

As for UX, it can help reducing bundle sizes (call heavy deps on the server and pass the output as props). Also, it can help solving the client the over-fetching problem that GraphQL solves, bc you can fetch data on the server and pass down as props only what's needed. In this sense, it's like a BFF

Instead of frameworks having to implement their own version of "data loader function" that only runs on the server to avoid client-server waterfalls (Remix loaders, Next's getServerSideProps), now you have a framework-agnostic implementation. And we get this following the good ol' component model.

I'll try to not sound like a fan boy, but I do see big DX and UX benefits.

3. Now we’re in the interactivity multiverse, but I want this to execute on the server. Let’s open a portal back to the server multiverse. The code below will be sucked into the server multiverse. This is "use server".

2. Does the component need interactivity? Remember, we’re in the server multiverse. Let’s open a portal to the interactivity multiverse. The code below will be sucked into the client multiverse. This is "use client".

My mental model for why we use a directive to “mark” client components, but not the other way around (a directive on server components), is this:

1. We start off on the server multiverse—be it the CD server, a long-running Node.js server, etc.

I don't like at all that many people believe that React == Next.js, but then we all saw that server functions were announced in NextConf before any docs on that were written

both