spaceemotion

893 posts

spaceemotion banner
spaceemotion

spaceemotion

@spaceemotion

Loves to read books. Merges code with passion for design. Loves Soundtracks and makes stuff. Currently working on @novelcrafter - she/her

/dev/eth0 Katılım Eylül 2011
199 Takip Edilen109 Takipçiler
spaceemotion
spaceemotion@spaceemotion·
@wiedymi Whoops 😅 tried it with a variable font i use - the .ttf loads fine, but the .woff2 triggers "Symbol overflow". The TTF axes are also not working (in Photoshop it's fine), but Figma has the same issue with this one; it's a paid font so not sure how to best report this on GH...
English
2
0
1
27
Wiedy Mi
Wiedy Mi@wiedymi·
I use text-shaper that i built few months ago for subtitle rendering to make this. Pure TypeScript text shaping engine. Extremely fast and accurate. Zero dependencies.
Cheng Lou@_chenglou

My dear front-end developers (and anyone who’s interested in the future of interfaces): I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept): Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow

English
4
4
40
3.8K
spaceemotion
spaceemotion@spaceemotion·
@wiedymi I also wonder if Vercel's Satori could benefit of this - they do depend on the opentype.js library after all (we use them heavily for our own OpenGraph images)
English
2
0
1
17
spaceemotion
spaceemotion@spaceemotion·
@wiedymi The benchmarks in the README look great - any chance of seeing how it compares to pretext? (or have a section comparing your library to theirs?) it looks like most of the benefit is for server-side-browserless rendering of text to SVG/images, right?
English
2
0
2
40
spaceemotion
spaceemotion@spaceemotion·
@bdc CSS is amazing, but i too have hit many performance ceilings where i hit the ceiling due to layout + repainting (even with CSS containments and such). I personally see this very useful in canvas/whiteboard tools like @tldraw where every ns matters.
English
0
0
3
1.1K
Benjamin De Cock
Benjamin De Cock@bdc·
Pretext is awesome for other reasons, but seeing all the comments on "editorial design being finally possible on the web" makes it clear most people don't know that ˋshape-outsideˋ does the job natively in CSS for DOM layouts in all browsers.
Cheng Lou@_chenglou

My dear front-end developers (and anyone who’s interested in the future of interfaces): I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept): Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow

English
8
5
143
23K
spaceemotion
spaceemotion@spaceemotion·
@itaydre @Affinity You could already use the `shape-outside` CSS property (developer.mozilla.org/en-US/docs/Web…) to achieve this. afaik, the library is only for not having to wait for the browser to do the layout, and getting slowed down by the back and forth. it's more of a perf. optimization than capabilties.
English
0
0
1
67
spaceemotion
spaceemotion@spaceemotion·
@kettanaito is the video above showing you working on your own editor with the latest text layout library that everyone's been sharing?
English
1
0
0
43
Artem Zakharchenko
Artem Zakharchenko@kettanaito·
Rich text editors look awesome in the promo materials and then you start using them and the tiniest of things are falling apart by default 😔
English
4
0
22
3.1K
kepano
kepano@kepano·
Defuddle now returns Youtube transcripts! Paste a YouTube link into defuddle.md to get a markdown transcript with timestamps, chapters, and pretty good diarization! ...or if you just want to read it, try the new Reader mode in Obsidian Web Clipper powered by Defuddle.
English
90
199
2.7K
327.9K
spaceemotion
spaceemotion@spaceemotion·
@astuyve @khajimatov @JackEllis a lot of the requests are just simple writes, though we do use snappy and brotli compression a lot (e.g. right now it takes around 16sec to load about 180MB of data from dynamodb + uncompress + serialize to JSON + compress to brotli for the response.)
English
0
0
1
21
AJ Stuyvenberg
AJ Stuyvenberg@astuyve·
Lambda just got beefier – up to 32gb RAM and 16 vCPU if you're in LMI mode (serverless servers)
AJ Stuyvenberg tweet media
English
6
1
32
5.6K
spaceemotion
spaceemotion@spaceemotion·
@astuyve @khajimatov @JackEllis hm. well we are pushing around 50 mill. req. per month (~30 req/s) through lambda. so far it's been not an issue though (also got a couple reserved instances). so nothing we'd need a fleet of servers for, really.
English
1
0
1
76
AJ Stuyvenberg
AJ Stuyvenberg@astuyve·
@spaceemotion @khajimatov @JackEllis LMI is a lot different operationally than those, and is primarily a way for heavy users to maximize their cpu time. I’m not sure what the break even point is but I’d imagine it would need to be close to 5-10k per month for it to really be worthwhile
English
1
0
1
98
spaceemotion
spaceemotion@spaceemotion·
@astuyve @khajimatov @JackEllis Wait - i totally missed that then. I saw the announcement and thought "prob. not needing it". Why did AWS not market it as "Fluid" compute-ish (or at least compare it to Vercel or Cloudflare workers and such?) I also always wondered what to do if a request kills the whole server.
English
1
0
1
139
AJ Stuyvenberg
AJ Stuyvenberg@astuyve·
@khajimatov @JackEllis vercel built there own version, it shipped before AWS did this. I think vercel's is actually more useful because they handle all of it for you. AWS's version will probably save you more $$$
English
3
0
1
162
spaceemotion
spaceemotion@spaceemotion·
@DanielLockyer I recently ran benchmarks to compare marked, markdown-it and micromark - and was quite shocked at how bad micromark performed, given that so many packages build upon it.
spaceemotion tweet media
English
0
0
1
46
Daniel Lockyer
Daniel Lockyer@DanielLockyer·
I would love to see more nerd sniping in the performance arena
English
4
1
29
2K
spaceemotion
spaceemotion@spaceemotion·
@thdxr I stopped using it because the git tracking slowed things down! Once I disabled it, the performance went right back up.
English
0
0
0
80
dax
dax@thdxr·
do you use the opencode sidebar
English
84
0
81
25.5K
spaceemotion
spaceemotion@spaceemotion·
@DanielLockyer Felt very proud when we got this comment on our software - a lot of care has been put into making @vuejs even snappier and faster than it usually is (the very fine grained reactivity it allows really helps!)
spaceemotion tweet media
English
0
0
1
58
Daniel Lockyer
Daniel Lockyer@DanielLockyer·
I wish I had the time to fix all these performance issues I come across, but the reality is that I don't I hope to inspire more people to look into the speed of the code they use and produce, and maybe one day we'll be living in a faster world 🧘🏼
English
9
0
45
3.2K
spaceemotion
spaceemotion@spaceemotion·
@DanielLockyer Really happy that OTel is getting a closer look at performance! We've actually been thinking about dropping it in favor of custom, specific metrics due to all the overhead so far...
English
0
0
1
332
VJ
VJ@andrewvijay·
Vite 8 is faster than v7 but it's still slower than rsbuild. 3.5s vs 1.2s
English
2
3
115
47.2K
spaceemotion
spaceemotion@spaceemotion·
@youyuxi @andrewvijay I had this option "break" our build with the AWS SDK v3 - something in the @ smithy packages breaking, even though they're marked as sideEffects: false. Our post-build tests say it can't find an enum in there: #L7" target="_blank" rel="nofollow noopener">github.com/smithy-lang/sm…
English
1
0
1
182
Evan You
Evan You@youyuxi·
@andrewvijay I think it’s simply because we haven’t enabled lazy barrel optimization by default yet. It’s still behind a flag and I assume rsbuild does it by default. Try turning it on via build.rolldownOptions.experimental.lazyBarrel: true #configuration" target="_blank" rel="nofollow noopener">rolldown.rs/in-depth/lazy-…
English
2
0
14
2.7K
spaceemotion
spaceemotion@spaceemotion·
@sam_paech I don't think the Hunter/Healer stealth models are actually Qwen - for us they perform worse than even the open weight models that are already out. They're closer to GLM in performance in that way 🤔
English
0
0
3
204
Sam Paech
Sam Paech@sam_paech·
New results! GPT-5.4: Places 1st, 2nd & 3rd on creative writing, longform writing & EQ-Bench respectively Grok-4.20: Refusals everywhere, and consequently low scores Hunter-alpha: New 1M context stealth model on openrouter. Possibly Qwen-3.5 max?
Sam Paech tweet mediaSam Paech tweet mediaSam Paech tweet mediaSam Paech tweet media
English
39
42
468
56.1K
spaceemotion
spaceemotion@spaceemotion·
@sam_paech @flopsy42 @altryne Can confirm from our own benchmarks - reasoning does not improve the creative writing performance of GPT-5.4. It does, however, help with text rewriting tasks, or reducing hallucinations (e.g. evals that check for false positives)
English
0
0
3
73
Sam Paech
Sam Paech@sam_paech·
@flopsy42 @altryne Reasoning off for all of these evals, where possible. I find it never helps and usually harms performance on these tasks.
English
1
0
4
418
spaceemotion
spaceemotion@spaceemotion·
@mattpocockuk Being able to whip up quick personal tools without needing to google for a cloud-based solution that doesn't quite does the thing, is great (and no need to spend time to write it on your own)! tons of little gadgets that help in daily work :)
English
0
0
1
151
Matt Pocock
Matt Pocock@mattpocockuk·
I'm collecting stories from folks who've had a genuine 'this changes everything' moment with AI coding. What was yours?
English
271
13
269
70.5K

Keşfet

@wiedymi @bdc @tldraw @itaydre @Affinity @kettanaito @kepano @philmohun