OSCity

Today I launch the website of @ArchiyouApp at archiyou.com. A cloud platform for designing with a code language so simple that I hope it can introduce designers to coding and programmers to design.

Experimenting with procedural generated words in @threejs

Video diffusion models learn motion indirectly through pixels. But motion itself is much lower-dimensional. We introduce 64× temporally compressed motion embeddings that directly capture scene dynamics. This enables efficient planning -> 10,000× faster than video models. 🧵👇

This border fence is not for people. It was built along the Finland–Norway border to keep reindeer from crossing, while humans use the steps to pass over it.

My team at Iwfgara(iwfgara.com) has successfully visualized FourCastNet3 AI weather models from #NVIDIA Earth2Studio. The synergy between AI-driven NetCDF data and our 3D engine is stunning. The future of digital twins is here.

Introducing Kernels on the Hugging Face Hub ✨ What if shipping a GPU kernel was as easy as pushing a model? - Pre-compiled for your exact GPU, PyTorch & OS - Multiple kernel versions coexist in one process - torch.compile compatible - 1.7x–2.5x speedups over PyTorch baselines

Mini-engine with render thread and new render data model (running on Metal). WebGPU will be running in WASM worker, presenting from there. The main thread is still doing culling. Will move that to render thread next.

Spark 2.0 is here! 🚀 We’re redefining what’s possible on the web with a streamable LoD system for 3D Gaussian Splatting. Built on Three.js, you can now stream massive 100M+ splat worlds to any device from mobile to VR using WebGL2. All open-source. Dive into the tech 👇

Introducing Neuform Generate next-level designs with a living DESIGN.md. Turn design systems into skills and remix into any format with unique animations, webgl and threejs. It's available on neuform.ai

This 15-minute talk by the creator of Pydantic on how to correctly use MCPs will teach you more about making your AI tools actually work together than everything you've scrolled past this year. Bookmark this & watch, no matter what. Then read the guide below by @eng_khairallah1

bro was right. Atlassian down 75%. HubSpot down 69%. Figma down 86%. Almost all of them down 30–70% from their 52-week highs. AI is literally eating software alive and repricing every company in real time. SaaS is cooked fr 😭

NEW paper from Meta. (bookmark this one) What if the model wasn't just using the computer, but became the computer? New research from Meta AI and KAUST makes a serious case for Neural Computers (NCs). The paper proposes NCs as learned runtimes where computation, memory, and I/O live inside a single latent state. Their first prototypes use video models to roll out terminal and GUI interfaces from prompts, pixels, and user actions. Why does it matter? Today's agents still depend on external computers to store state, execute actions, and enforce system contracts. Neural Computers point to a different machine form: one where interface dynamics, working memory, and execution are learned together. The early results are promising but grounded. CLI rendering improves, GUI cursor control reaches 98.7% with explicit visual supervision, and reprompting boosts arithmetic-probe accuracy from 4% to 83%. But symbolic reliability, stable reuse, and runtime governance remain open. This is less "agents got better" and more "what comes after agents as a computing substrate?" Paper: arxiv.org/abs/2604.06425 Learn to build effective AI agents in our academy: academy.dair.ai

I was thinking about the evolution of knowledge formats in the digital era and how AI fits into this history - and made this diagram. Your comments are welcome!

Mapping an entire country’s agriculture typically requires massive computational resources. However, researchers have just mapped one country's cropland in 16 hours for only $313.40. Here's the breakdown:

カラクリストさんの個展 「可視化する機巧II」 凄い！（語彙力）

西八王子のギャラリーいちょうの木で開かれている、からくりすとさんの個展「可視化する技巧II」を拝観しました。生で見ると技巧が凄いですね…。

Demis Hassabis just described the moment medicine stops treating disease and starts deleting it from the source code. For all of human history, doctors have fought symptoms. The tumor. The organ failure. The collapse. Always downstream. Always after the damage has already started. The cause sits upstream. Written into the DNA. Ninety-eight percent of the human genome sits in non-coding regions. For decades, science understood the genes but couldn’t read the vast dark territory between them. That’s where most disease hides. Hassabis: “It takes the big, long genetic sequences and then it tries to predict, if you made a mutation to this particular single letter, single position in the genetic sequence, will that be a harmful mutation that might cause disease, or is it benign?” AlphaGenome reads your entire genetic sequence and identifies the exact letter that’s corrupted. Not a region. Not a probability range. A single position in a three-billion-letter sequence. That alone would be a generational breakthrough. But most diseases aren’t that clean. Hassabis: “What if they’re multigenic diseases where there’s cascades of mutations causing the problem? Those are even harder to detect, but actually perfect for sort of AI.” One mutation is hard enough to find. A cascade of mutations interacting across the genome is a problem no human researcher can hold in their head at once. Three billion data points. Compounding errors across all of them. The human brain cannot solve that. AI doesn’t solve it either. It maps it. All of it. At once. The most devastating diseases on Earth. The ones medicine has called untreatable for generations. They are not mysteries to the algorithm. They’re compute problems. But finding the error was only ever half the equation. You also need the ability to fix it. That tool already exists. CRISPR is a molecular scalpel. It cuts DNA at exact positions. The limitation was never the editing. It was knowing exactly where to cut. Hassabis: “A kind of combination of things like AlphaGenome and CRISPR could be incredibly powerful.” AI reads the code. CRISPR rewrites it. One finds the mutation. The other corrects it at the source. Not managing symptoms. Not slowing progression. Deleting the error from the genome. The implications go beyond treatment. A disease corrected at the genetic level doesn’t just disappear from one patient. It disappears from their bloodline. The read access is here. The write access exists. The merge is inevitable. The era of accepting a broken genetic hand is ending. We stopped being passengers in our own biology.

We've been tricked, again. Many of the thousands of bugs and vulnerabilities Mythos found are in older software are impossible to exploit. And the severe zero-day reports rely on just 198 manual reviews tomshardware.com/tech-industry/…

I thought that graphics engineers were bad at standardizing. AI engineers seem even worse :D

Psychologists have posited hundreds of cognitive biases over the years. A fascinating new paper argues that they all boil down to one of a handful of fundamental beliefs coupled with confirmation bias. stevestewartwilliams.com/p/one-bias-to-…

I made a quick little threejs previewer for patina outputs. paste the json, click apply, preview it within threejs on a plane or sphere with an HDRI or basic lighting (via the Three Lights option in the HDRI dropdown) patina-preview.vibe-coded.com