GPUI's State Model Fails to Scale
GPUI and egui state models can't say what changed, so applications pay for full re-renders. A label update in a UI of thousands of rows takes 8.4 ms in GPUI. — blog.zortax.de
PixelOS Brings Pixel Experience to AOSP
PixelOS is an AOSP-based ROM that bundles Google apps and Pixel features with performance enhancements. — pixelos.net
Chimera Simulates Vocal Tract Physics
Chimera produces every sound by simulating air moving through a virtual throat, mouth, and nose, letting users manipulate the tongue, lips, and jaw. — chimera-vocal-synth.vercel.app
Dux Modular Shell Releases Stormborn Update
Dux Modular Shell 2.1 Stormborn adds a new system monitor, video wallpapers, and a revamped media player to its desktop environment. — duxlabs.dev
Crowdsourcing Pangram Scans Divides Lobsters
A shared index of Pangram scans for Lobsters is gaining interest, but many users reject it because false positives can brand human writing as AI-generated. — lobste.rs
TinyLittleOffice Shows Web's Power
TinyLittleOffice, a web office suite that stores data in the URL, drew hundreds of thousands of visitors and showcased the browser's low-cost app-building power. — ftrain.com
Uncensored Gemma Fine-Tune Exceeds Bigger Base
A fine-tune of Gemma 4's base model activates only 4.5 billion parameters while exceeding the larger Gemma 4 model in critical benchmarks. — huggingface.co
RAM Altman Script Mocks OpenAI CEO
A developer's BASH script emails Sam Altman sarcastic messages about idle RAM, mocking his statements about memory shortages. — respawnfirst.com
Asahi Linux Adopts GPL to Stop Code Laundering
Asahi Linux relicenses all its code under GPLv2 or later after unreleased work was lifted from developers' forks, often with AI used to hide its origin. — itsfoss.com
Qwen3 Fine-Tune Offers Lightweight Local Chat
A Qwen3 fine-tune activates only 4 billion parameters per token, making it a lightweight GGUF choice for local chat. — huggingface.co