voronoids - voronoi cells using boids as seed points, + CRT shader.
threejs / TSL / webgpu
voronoids.vibe-coded.com
voronoids - voronoi cells using boids as seed points, + CRT shader.
threejs / TSL / webgpu
voronoids.vibe-coded.com
We need trees of PRs in Github.
passing on the good word:
"Threejs Conf Paris, here we go ! π --> threejs.paris
Waitlist open !"
organized by David Ronai (et al.), already a solid lineup :)
In a HN comment someone used the term "pilot" instead of "author" for AI-generated code.
Sounds appropriate.
Ran into some issues with jj workspaces. It seems you have to be careful about what workspace you are in when manipulating the changes graph
Pure talent.
To clarify: AI is now good enough to lower the barrier to entry to do exactly that.
But in practice the people who build and ship things are those who didn't wait for AI to build and ship things.
"AI allows me to build things I would never have the time to build before"
The things:
You missed the "It doesn't work for me"
I am way less tempted to consult social media and news websites when I reframe them as "the voices in my head"
"Revenge on the Nerds"
10m documentary film on the local-first software movement!
www.youtube.com/watch?v=10d8...
What I'm saying is, you made you non-work life too comfortable.
Organize a conference.
#grow2026
Went through the same thing 2 years ago. But is it really the market or did we simply isolate ourselves?
Less in Paris, less social media presence, staying longer at the same companies...
Maybe we just stopped doing the things that made our (work) lives easier 10 years ago
Been ruminating on this lately as a maker of small one-off thingsβ¦ years back Iβd made console.frog github.com/tholman/cons... β¦ prob took a few hours, would take ai 1 minute. Itβs not that my enjoyment from executing on an idea is gone, but the way my work is perceived has shifted.
At work things are different, these discussions still happen, just not on slack, more in small informal calls between developers
Just yesterday I was wondering why the "webmestre" slack was so quiet these days.
Some hypothesis:
- LLMs, at least for some channels.
- tech stabilization: there is less to say about front-end frameworks than a few years ago
- us getting older, which can have an impact for many reasons
"Order and disorder." probably my last take at #genuary26, uses a 3D pathfinder to go from neatly ordered stacks to a random layout. #genuary2615
BTW, I'm available for freelance work (2 days a week for now).
HIRE ME, I'm GOOD (& CHEAP! π)
also ok with collabs, mentoring, Q&As, AMA...
Friendly reminder that if you're doing #genuary with either 2D canvas, p5.js or three.js, I would love for you to try out Fragment
github.com/raphaelameau...
#genuary2026 #madewithfragment
I had Steve Yegge's Gastown in mind not the crypto stuff.
I find the confusion ironic because I think I used his own words.
For many of us there is a gap between what we read online and what we can achieve with AI.
This is an issue at work because management read a lot of LinkedIn.
Having a serious conversation about how to build reliable software with AI, without having to review 10,000 LoC AI-slop PRs is welcome.
It's ok as long as you are reasonable and don't try to sell snake oil.
I doubt gastown "it makes me 10000x more productive, use at your own perils, yolo, btw it costs $20,000 a day" would be welcome here
Circles, feedback loop
It's not really related to matrices, cross and dot products are related to cos and sin, a direction can be a normalized vector instead of an angle
Just a different way of looking at things
Inigo Quilez wrote multiple "avoiding trigonometry" articles:
iquilezles.org/articles/noa...
iquilezles.org/articles/sin...
iquilezles.org/articles/noa...
I remember reading the first one years ago, it completely changed how I think about angles.
The most likely outcome is that we will completely stop doing code reviews
Oh, and avoiding trig functions. One acos() made it through, but in general, if you use any trig functions in this, you're doing it wrong and there's almost certainly a better way. That's some algebra I'm glad I didn't have to repeat.
I haven't tried TSL, but I know there is this vite plugin to help with the operators: github.com/makio64/vite...
My best suggestion is pair (not peer) review: the "author" and the reviewer look at the code together.
Ensures the author has read the code at least once and the reviewer doesn't have to face alone AI slop nobody took the time to understand.