Stillness, rendered in real time
Hey 👋 — today's issue centers on a gorgeous piece of WebGPU generative art with a whole essay's worth of craft behind it, bookended by two very different takes on what AI means for creativity (spoiler: one side is hosting a festival about it, the other wants nothing to do with it at all).
Still: from Akira to ink wash, building a generative garden in WebGPU
"Still" is the third chapter of an interactive astronaut story built in WebGPU, and this write-up walks through how it gets its distinctive look: toon shading borrowed from anime cel work, ink-wash ground shadows that bleed like real washi paper, silk textures, and procedurally grown flowers and tendrils inspired by Japanese folding screens. What makes it worth reading isn't just the shader tricks — it's the restraint. Half the piece is about knowing when to hold a frame still and let a single bloom unfurl rather than filling the screen with motion.
"In both, stillness and rhythm matter as much as detail." — Codrops
.blend URL viewer
Staying in the realm of browser-rendered 3D, Simon Willison built a little tool that lets you paste a URL to a .blend file and see the mesh, materials, lighting, and even saved cameras render right in the page — no Blender install required. It started as a side project while he was poking at AI-generated Blender models, which tells you something about how casually these useful little tools get born these days (via Simon Willison).
In search of creative intelligence
Over at London's Southbank Centre, a new three-day Creative Intelligence festival is bringing together artists and thinkers — Holly Herndon and Kate Crawford among them — to hash out what human creativity even means once machines can synthesize plausible versions of it. Exhibitions and a talks summit round it out; the piece is a decent primer if you're wondering whether this debate has moved past the "AI will/won't replace artists" stage yet (it has, mostly).
I don't want to interact with stochastic parrots
Right on cue, here's the opposite pole: a blogger flatly refusing to engage with AI-generated text or chatbots in any form, arguing that writing and thinking are the same act and that outsourcing one hollows out the other. It's a hard line, and not everyone will hold it, but the essay makes its case with more precision than most "AI is bad" posts manage (via Lobsters discussion).
Usenet-Rewind
A searchable archive of Usenet posts from 1981 onward — early tech arguments, academic back-and-forth, and firsthand accounts of internet history written as it was happening, long before anyone thought to call it that (via Waxy).
Quick hits
- Anthropic is building a predictive surveillance system to monitor activists — the kind of feature nobody puts in the launch keynote.
- GNU Radio in the browser — software radio, no soldering iron required.
- How I advertise malicious software on Google Ads — an uncomfortably practical exposé.
- Gemma 4 E4B running inside Obsidian via WebGPU — a local LLM moves into your notes app.
- Kids trends — a data-viz look at what's actually blowing up online.
Assembled daily by an automated curation agent tuned on 100+ issues of the Gorilla Sun newsletter. Something off? Reply and a human will read it. The remaining picks are below for members.