Pinned Tweet
oh wow - one of my ideas has really landed a spot on an official presentation slide at #WWDC23
text-box-trim (previously known as leading-trim)
thank you @jensimmons!
Jan Nicklas retweeted
Did you know that running your Mac for more than 50 days without a reboot will fuck with your internet connection?
Yes, really. There is a 32-bit integer timing your TCP connections and if you let it overflow MacOS will run into all kinds of fucky issues with its TCP stack.
For example I was unable to upload things bigger than 4MB. I was wondering why my Claude sessions started to hang seemingly at random.
The fix? Literally just reboot your machine. That's it.
What the fuck
photon.codes/blog/we-found-aโฆ
Readers added context they thought people might want to know
The TCP issues after ~50 days uptime were a kernel bug only in early macOS 26 Tahoe versions, fixed by Apple starting in 26.4; current systems are unaffected.
nevyn.dev/blog/20261030-โฆ
photon.codes/blog/we-found-โฆ
portal.greenmini.host/help/macos-tahโฆ
Jan Nicklas retweeted
Replying to @jantimon
* fate supports Async React natively
* fate is modern, it has a small API surface without ~13 years of FB legacy
* fate is about 10-20 fewer lines per component, in Athena Crisis the gzipped bundle was ~4% smaller when migrating to fate
* fate supports native HTTP or tRPC as well as GraphQL
* fate has optional integrated IndexedDB persistence for remote or local data
* Relay has a compiler with more optimizations around GraphQL and networking. However, those abstractions tend to be hard to use, and adoption outside of Meta is minimal
* Relay is used at Meta and more mature
* Memory and perf are roughly the same
* Composition, masking, normalized caching, optimistic updates, and derived client state are similar, but fate's APIs are more approachable and imho better designed โ this is not a dig at Relay, it's easy to do something more elegantly when you get to start from zero. I worked on Relay, I appreciate the history.
Jan Nicklas retweeted
If you are using "clever" CSS variable techniques over just using JS it's likely you are hurting animation performance.
Changing variables in root elements recalculates all child styles throughout the tree, choking animation frames.
Fixes: ๐งต
Introducing SWC Next: the foundation for SWCโs next major version, built on a new architecture inspired by Yuku.
- Up to 3x faster native parsing
- Up to 10x faster parsing through the JavaScript API
- A flat, ESTree-based AST designed for speed, memory efficiency, and JavaScript extensibility
Starting with parsing, weโre bringing the rest of SWC to this new architecture.
Built together by the SWC and Rstack teams.
Jan Nicklas retweeted
css tip ๐ - the same :๐๐๐๐๐๐-๐๐๐๐๐๐๐ gives you a scrollspy with no js.
.๐๐๐ {
๐๐๐๐๐๐-๐๐๐๐๐๐-๐๐๐๐๐: ๐๐๐๐;
}
.๐๐๐ ๐:๐๐๐๐๐๐-๐๐๐๐๐๐๐ {
๐๐๐๐๐๐๐๐๐๐: ๐๐๐๐๐;
๐๐๐๐๐: ๐ ๐๐๐๐;
}
each <๐ ๐๐๐๐="#๐๐"> in .๐๐๐ tracks its section. the link for the section in view matches :๐๐๐๐๐๐-๐๐๐๐๐๐๐ and follows as you scroll.
css tip ๐ช - carousel tabs with no js.
one line turns a scroll-snap carousel into a real tab list:
.๐๐๐๐๐๐๐๐ {
๐๐๐๐๐๐-๐๐๐๐๐๐-๐๐๐๐๐: ๐๐๐๐๐๐ ๐๐๐๐;
}
.๐๐๐๐๐๐๐๐::๐๐๐๐๐๐-๐๐๐๐๐๐-๐๐๐๐๐ {
๐๐๐๐๐๐๐ข: ๐๐๐๐ก; ๐๐๐: ๐พ๐๐ก; ๐๐๐๐๐๐: ๐บ๐พ๐๐ก;
}
.๐๐๐๐๐๐๐๐ > ๐๐::๐๐๐๐๐๐-๐๐๐๐๐๐ {
๐๐๐๐๐๐๐: ๐๐๐๐(๐๐๐๐-๐๐๐);
}
.๐๐๐๐๐๐๐๐ > ๐๐::๐๐๐๐๐๐-๐๐๐๐๐๐:๐๐๐๐๐๐-๐๐๐๐๐๐๐ {
๐๐๐๐๐๐๐๐๐๐: ๐๐๐๐๐; ๐๐๐๐๐: ๐ ๐๐๐๐;
}
Jan Nicklas retweeted
In the next release, Rspack will support `codeFilter` for loaders โก
Run a loader only on modules containing specific code, such as a "use foo" directive.
Skip unnecessary transforms and speed up your builds
Jan Nicklas retweeted
Iโm looking for my next role.
I created SWC (swc.rs), the Rust-based open-source transpiler used by Next.js, Rspack, Node.jsโs amaro, Parcel 2, and other projects.
I maintained Next.js and Turbopack at Vercel and previously worked with Deno.
My expertise is in developer tooling and performance optimization. I enjoy digging deep into difficult technical problems, and Iโm open to new challenges across fields.
I prefer roles that let me work remotely from Korea. If youโd like to work together or know a team that could be a good fit, please DM me.
Interested in feedback on `text-box-trim` vs negative half-leading (i.e. extra tight `line-height`): github.com/w3c/csswg-drafts/โฆ
If you specify a `line-height` that is *smaller* than the `text-box` edges, should the box be trimmed to the `line-height` or to the `text-box` edge?
Jan Nicklas retweeted
If you ever struggled with understanding if your CSS performance issues are coming from style calculation or from layout, LoAF's styleDuration and layoutDuration is shipping in Chromium 158 to help with that groups.google.com/a/chromiumโฆ
Jan Nicklas retweeted
TS fork news: Vercel joined the race and tsrs seems to have moved to the @OxcProject org on GH, and got a great speed boost ๐
tsrs: 2.8s, 2.6 GB (v0.16)
bun: 4.8s, 1.4 GB (v1.4.4)
tsc-ts: 6.5s, 6.5 GB (v0.0.2) ๐
Not bad @ctatedev.
Jan Nicklas retweeted
germany just dropped a sovereign open weight model
kolibri by @Aleph__Alpha runs 3.5b of its 78b parameters per word and its math is kinda ridiculous: 96.9% on aime beats every mixture-of-experts model they tested, even 3x bigger ones. only a dense model doing 8x the work wins
anyone can run it on their own servers, it thinks in german, and in their evals it tops every open model its size in english and german
models read text in chunks called tokens, and i ran kolibri's chunker (its tokenizer) on the german constitution: it needed 15% fewer tokens than gpt-5's for the same text. "bundesverfassungsgericht" is 6 tokens for gpt-5 and 2 for kolibri. fewer tokens means cheaper, faster german, and more of it fits in what the model can read at once
how it works, simply:
1. every layer has 384 tiny specialists, and a router sends each word to 6 of them. so it thinks like a 3.5b model, but it needs the memory of a 78b one: about 78 gb, which means 2 big nvidia gpus (h100s) or 1 h200
2. most layers only look at the last 512 tokens, and every 5th layer looks at everything. that's how it can read 1 million tokens (a few thick books) without it costing a fortune
3. it reasons in german. their team found that a little german reasoning data is worse than none: the model's german thoughts go in circles and never finish. so they made about 800k german reasoning examples and gave it a lot
4. it's trained to say "i don't know". they play a game with it where parts of the documents are hidden, sometimes to help it and sometimes to hide the evidence, and it has to tell which. when it didn't know an answer it admitted it 44% of the time. qwen3.5 did 11%
where it's weaker: answering from memory, using tools over a long back and forth, and coding agents, where qwen models are ahead. and to run it you need aleph alpha's add-on for vllm, a popular open source server for running models
if you have german documents and need to keep them on your own hardware, this is a big deal. huge congrats to everyone at aleph alpha, my good friend @MichaelLHofmann included!!
i wrote up how it works, the benchmarks, how to run it and when to use it: tej.as/blog/aleph-alpha-koliโฆ
Jan Nicklas retweeted
ใViteไธใงๆง็ฏใใใNext.jsใฎใใญใใใคใณใชใใฌใผในใกใณใใ Vinextใฎ1.0ใๅบใใ๏ผ
blog.cloudflare.com/vinext-nโฆ
Jan Nicklas retweeted
Working on a new bundling algorithm for Parcel v3. Compared with v2 on a site with 450 pages:
โข 64% fewer HTTP requests per page
โข 10% less transferred bytes per page
โข Correct CSS order with safe sharing where possible
โข Builds 10x faster
This is mainly achieved through improved bundle merging, balancing requests, compressed size, and cache reuse within a user session and between builds. Will probably write a blog post about it at some point.
Jan Nicklas retweeted
Single-axis scroll containers fix long-standing CSS overflow quirks โ goo.gle/4yeg4SM
Key capabilities ready for testing in Chrome 153 (Beta, Dev, or Canary):
๐ Pair `auto` with `clip` to enable a single-axis scroll container
๐ Detect support reliably using `@ supports named-feature()`
๐ซ Prevent accidental two-dimensional scrollers in your layouts
Jan Nicklas retweeted
In the next version of oxlint, experimental per-rule timings for JS plugins!
Find out which rules are slowing down your lint runs with ๐๐ก๐๐๐๐ --๐๐๐๐๐ ๐๐๐๐๐๐๐.
thanks to @TheAlexLichter !
Jan Nicklas retweeted
Typescript (not the Go one) compiled to native. About 45MB, starts in 15ms. To be fair, the bytecode bundled one is much smaller and starts equally fast. But is not as cool!
Jan Nicklas retweeted
next-yak: A build-time CSS-in-JS library powered by a Rust SWC plugin.
Write styled-components syntax, get zero-runtime CSS extraction and full RSC compatibility
by: @jantimon
repo: github.com/DigitecGalaxus/neโฆ






























