ElleMeDit retweeted
오늘 영광스럽게도 유신 교수님의 초청으로 KAIST CS350 Introduction to Software Engineering 과목에서 특강을 하게 됐는데요, 강의 자료를 공유합니다: Nobody's job, everybody's problem: F/OSS in the age of AI.
hongminhee.codeberg.page/fos…
ElleMeDit retweeted
파운데이션 파트너의 글. 조직이 새로운 것을 시도할지, 지금처럼 할지에 대해 새로운 것을 탐색하지 않은 진짜 이유는 비용 때문 만은 아니다. 실패했을 때 잃는 게 더 많은 사람들이 결정권을 갖고 있었기 때문이다. cms.digitaltoday.co.kr/news/…
I’ll die on this hill: “isomorphic” means it’s 1-1 and onto, and SSR is neither. And hydration isn’t invertable or structure preserving either. So nothing about SSR is isomorphic (though I still call it that too).
ElleMeDit retweeted
The security vulnerabilities in React Flight a few months ago were indeed serious and it's fair to scrutinize how they happened, but if you do, I don't think any fair person would attribute the cause to "sloppiness" or lack of rigor. A huge amount of care went into designing that protocol over many years.
I've come to the sad conclusion that @apply in Tailwind is a big anti-pattern.
With enough complex utilities, your CSS can balloon to hundreds of kB, bloating the CSSOM and hurting runtime rendering performance, not just network.
It’s a bummer because I was really excited to use the new @utility feature to build framework-agnostic components, but it doesn’t scale well.
Best approach is inline utilities as usual, plus a CVA-like library when you need reuse. I’m working on something to make that easier.
ElleMeDit retweeted
We cut @nodejs memory in half with a one-line Docker image swap.
No code changes. No new APIs. Just smaller heaps.
Here's what happened when we tested pointer compression on real workloads 🧵
ElleMeDit retweeted
동료 분과 요즘 자주했던 얘기지만, 생산성을 올리는 거 자체가 중요한 건 아님. 당연한 게 아무런 수요가 없는 물건을 10배 더 찍으면 재고 부담만 떠앉는 일임. 거꾸로 말하면 생산성 증가를 돈을 바꿀 수 있는 공식을 찾아야하는데, 이 때 생산성 증가를 무제한으로 받아줄 수 있는 곳은 어디인가.
ElleMeDit retweeted
Stacked Diffs on @GitHub will start rolling out to early design partners in an alpha next month. In the meantime, here's video of our progress so far:
(h/t for @georgebrock + team for their awesome work)
ElleMeDit retweeted
Pro tip for Claude Code: Add this line to your system prompt:
"Your context window will be automatically compacted as it approaches its limit. Never stop tasks early due to token budget concerns. Always complete tasks fully, even if the end of your budget is approaching."
Claude's context awareness makes it conservative about finishing tasks. This prompt unlocks persistent, autonomous execution across unlimited context.
platform.claude.com/docs/en/…
ElleMeDit retweeted
Spoiler: for the problems I list, it's not.
Any CTO or VP Eng who's tackled hiring or scaling JS teams over the years will tell you the same. You're taking on a fundamentally different runtime, return types, and control flow from idiomatic JS. You're taking on A LOT.
However...
ElleMeDit retweeted
TLDR: It frees up to megabytes of RAM for some real world Next.js App Router apps, especially content heavy sites.
Next.js now clears the RSC payload from browser memory after hydration.
Great work by @shuding
ElleMeDit retweeted
I've made a PR to optimize some of that work. It currently creates 1745 ReadableStreams per request unnecessarily 😬
This fix eliminates that and you get a 10-38% perf boost (Deno users get the most!)
github.com/vercel/next.js/pu…
ElleMeDit retweeted
Replying to @RyanCarniato
Interesting read, @RyanCarniato. After writing (and deploying) an implementation of SSR and RSC in OCaml, I have a few thoughts.
Single language is just superior, yes, a bunch of people might want to use go/java/rust for their server actions but a) its a matter of implementing the sc model/streaming/suspense logic (very costly and high maintainability) and b) moving logic from/to backend/client becomes a nightmare (both rewrite and many worries)
In my case we have an isomorphic app from the start, because our build system is rigid
the shared modules (almost all of them) can be flagged with client/server components via ppx (a very fancy babel plugins). It's a clear layer for preprocessing components (react.client.component/react.async.component) and integrate with a backend fw (react.action)
when dealing with other languages, data types are an unsolved problem here. In our case, we use a mature ocaml-to-javascript compiler and the semantics for ocaml data and js data are very similar but I imagine not being the case for many langs (since we are talking about isomorphism, not boundaries where json mostly solves this)
ElleMeDit retweeted
Replying to @rickyfm
My only complaint here is that this compromise should’ve shipped years ago. I understand that it doesn’t feel great, but having nothing for so long sucked.
We’ve been making custom “useStableRef” and “useStableCallback” functions in user land for years.
ElleMeDit retweeted
The difficulty with all untrack APIs are that it's easy to opt-out in places you don't expect, creating bugs that are really hard to find.
Instead of exposing those issues and making React harder to use, we scoped useEffectEvent to a single component that's close to the Effect.
ElleMeDit retweeted
We just merged @nomsternom's work on deferred commit into the @htmlstandard navigation API: github.com/whatwg/html/pull/….
See github.com/wicg/navigation-a… for more detail. This lets you run async JS code before a navigation commits, and redirect if necessary.






















