Search
8 stories found
Discord's CTO says the desktop client is being rewritten in Rust with the GPUI framework, dropping Electron and React Native. An early Windows build was shown using about 290MB of memory, but there is no release date and no independent testing yet.
Colibri, a dependency-free C engine that keeps mixture-of-experts weights on NVMe and loads them on demand, is back on GitHub's daily trending list at about 47K stars. It really does run frontier-size open models on ordinary hardware, but independent tests show giant models crawl well below one token per second.
An OLED touch MacBook Pro with a Dynamic Island, M6 updates to the 14-inch MacBook Pro and iMac, and a $22,299 Mac Studio. Pre-orders are open; everything ships November 6.
AMD reported record Q3 2026 revenue of $13.4 billion, up 45%, as MI450 GPUs and Helios racks started shipping to Anthropic, Meta, Microsoft, OpenAI and Oracle. It forecasts about $17 billion for Q4, but free cash flow fell to $554 million as it built nearly $3 billion of inventory.
A CVSS 10 missing-authentication flaw in Siemens' SIMATIC IoT2050 Advanced gateways, CVE-2026-58115, is now being exploited, according to threat-intel firm VulnCheck. The attack is a single request to Node-RED's admin API, so any Node-RED instance left without a password is exposed in the same way.
Motorola's new flagship has a Snapdragon 8 Elite Extreme Gen 6, a 200MP periscope camera, a 7,000mAh battery and seven years of updates. Pre-orders open in the US on November 19. Its most unusual feature is official support for the privacy-focused GrapheneOS, which until now ran only on Google Pixels. Motorola hasn't said when that support arrives.
The MIT-licensed AiSOC ingests security telemetry, runs thousands of detection rules, has LLM agents investigate alerts and keeps a human in charge of any response. It runs on a laptop-class Docker setup with a small local model — here's what it does, what it needs and where it falls short.
MIT CSAIL researchers say OpenAI's GPT-6 Astra can identify what small networks do from their raw weights and edit a Qwen model to fix its math. The headline result, a jump from 7.9% to 48.1% on MATH-500, came from Python tooling and 50 forward passes, not from reading raw weights. Here's what each setup showed.
