the HTML in canvas github.com/WICG/html-in... proposal is both a long awaited bliss AND an absolute blasphemy. I'll keep an eye on it though :)

in the same spirit, have mixed feeling about pretext (github.com/chenglou/pre...) it looks pretty hacky but could prove super handy, need to try.

tried to count how many fucks he gave, too many.

at this point, you probably thought I'd abandon my cube right?!
ridiculous!
I was just busy adding sound FXs, a music player (loosely based on Otomata) but more importantly LIGHTNINGS !!!
wanted it to shriek "HMMETAAAAAL!🤘" in a high pitched voice when turning on but I failed...

haha true ! 😂
and, now that you mention it, "locally" could mean "they" live in a data center or something ... 🤔

"you can now run model blahblah locally!"

checks GPU requirements : 4 to 8 A100

"I don't think 'locally' means what you think it means..."

DLSS 5 to the rescue! (made with huggingface.co/spaces/victo...)

thank you 🙏
you do have a keen eye! 😀
it's a blue noise, it's fixed as it's used in the smoke to soften the raymarching steps but I guess it's OK to be fixed when you look THIS good right ?! (it's my fav noise ^^)

kept working on the 2.5D smoke thingy, it works by raymarching a 2D fluidsim and achieves interactive framerate on mobile (45-60fps) by reducing some settings.
it"s quite limited but allows some cool variations. video captured on a (5 year old) iPhone12 mini.
demo soon ™

currently testing Ace-Step for music generation ace-step.github.io/ace-step-v1....
fun but I guess one needs to be a musician to get the best of it.

it's crazy how unforgiving music is; the ears instantly catch AI slops (general weirdness, off beats, glitches) VS the eyes on genAI images & videos

art it is 👌
and flexing hard on the tech side too 😀
only checked the assets and some stuff look super tricky ( super smart too), they only load 2 videos (yoga and bite in the intro) all the rest is WebGL and so well executed.

me too, the whole concept hits hard 😂

ORYZO AI A physical product, reimagined for the AI era, proving even the smallest idea can feel extraordinary.

few people realize the level of craftmanship & involvement it takes to create a page like this oryzo.ai
by the @lusionltd.bsky.social team
not to mention the industry-grade disruption level of the AI product itself!

DLSS anything huggingface.co/spaces/victo... via @victormustar.bsky.social 🙏 (it uses FLUX.2-klein which I'll have to try at some point)

cool paper/repo of the day: lukashoel.github.io/video_to_wor...

reactions to Nvidia's DLSS 5 announcement really made my day 😂
bsky.app/search?q=DLS...
thank you internet 🫶

ooof ! amap-ml.github.io/RL3DEdit/ GS editing capabilities are getting serious

on the bright side, the design * successfully * passed all the slices !😀🤦

super nice! all of it :)

Screenshot of several images from my new blog post

Blog post: I've been working on a guide to SDF font rendering, and I kept getting stuck. The blog post is about the things that didn't work www.redblobgames.com/blog/2026-02...

yes :)

and yet another portfolio masterclass ✨👌
excellent work, as usual :)

Not sure what it is, but it is....
My new Portfolio! (best on desktop/laptop)
neuroproductions.be

Three.js Conf Paris The first Three.js conference in Paris. Join the community, discover the lineup, and be part of the event.

passing on the good word:

"Threejs Conf Paris, here we go ! 🚀 --> threejs.paris
Waitlist open !"

organized by David Ronai (et al.), already a solid lineup :)

spikes are definitely a must-have !
but first I need FIRE and LIGHTNINGS!
imagine spikes without fire and lightnings? preposterous.
... maybe some rain too ... for the drama 🤔
(thanks :))

No Rest for the Wicked Full Boss Fight - Caylen's Duo Plague Intensity 3 YouTube video by Melanator

thanks :) in the game, it looks like a "2.5D" smoke so my idea was to use a 2D fluid sim & the density as elevations then raymarch that volume.
smoke rendering resources : al-ro.github.io/projects/clo...
excellent article blog.maximeheckel.com/posts/real-t...
NRFTW
www.youtube.com/watch?v=uSwm...

thanks Kris, means a lot coming from you :)

Image Generation with a Sphere Encoder We introduce the Sphere Encoder, an efficient generative framework capable of producing images in a single forward pass and competing with many-step diffusion models using fewer than five steps. Our…

the project sphere-encoder.github.io was brought to my attention by Matt DesLauriers, I used Gemini to implement the paper and create a #WebGPU replacement for ONNX.js resulting in a lightweight inference engine.
repo: github.com/nicoptere/sp...
demo: barradeau.com/2026/sphere-...

I did a rogue version of the "sphere encoder", it's a smart way of embedding images that avoids having too large "gaps" between embeddings. it brings back the 2016 GAN vibe just that it runs at 60FPS in browser with a pure #WebGPU implementation now :) demo & code below

took me some time to realize "you" were not "the" David talking 😅it's packed with insightful tips & tricks, thanks for sharing!

How I made a shooter game in 64 KB YouTube video by daivuk

how a 64k game was made! What impressed me the most was the texture editing tool. youtu.be/qht68vFaa1M?...