2026-07-19
Top stories
- LG monitors install software and McAfee ads through Windows Update without consent LG monitors trigger Windows Update to install apps with McAfee subscription advertisements appearing automatically without user consent prompt.
- Moonshine Micro runs a full offline speech stack on an 80-cent microcontroller Moonshine AI published an offline voice stack with activity detection, command transcription, and text-to-speech on an 80-cent microcontroller.
- Gleam moves its canonical repository to Tangled The Gleam programming language moved its repository to Tangled, a decentralized social-coding platform built on AT Protocol like Bluesky.
Conferences and events
AI
Agentic coding
- Claude Code ships the Rust rewrite of Bun at scale Claude Code v2.1.181 embeds Bun v1.4.0, the Rust rewrite of the Bun runtime, making startup about 10 percent faster on Linux than Zig version.
- OpenAI cuts the bundled Codex model context window from 372k to 272k OpenAI reduced the bundled Codex model's context window from 372k to 272k tokens, cutting usable context by 27 percent with no model swap.
- Single-run comparison tests the /goal instruction on an NP-hard problem A blog post tested Claude Fable 5 and GPT-5.6 Sol on an NP-hard optimization problem with and without a time-boxed goal-directed instruction.
Security
Developer tools
- Transcribe.cpp offers a whisper.cpp drop-in with 60-plus ASR models Transcribe.cpp is a whisper.cpp drop-in supporting 60-plus speech-to-text models with acceleration backends for Vulkan, Metal, and CUDA.
- Minecraft Java Edition switches from GLFW to SDL3 Minecraft Java Edition Snapshot 26.3-4 replaces GLFW with SDL3 for window management and adds native Wayland support on Linux.
Hacker News
- Community tracker catalogs OpenAI Codex usage-limit resets An unofficial site called Codex Resets tracks OpenAI's unscheduled usage-limit resets, reporting 35 events with average 8.9-day intervals.
- Stack Overflow question volume charted against the arrival of AI assistants Stack Overflow question volume declined over recent years, with commenters attributing the fall to AI assistants and aggressive moderation.
- Blog post argues open-weight models have reached parity for everyday coding A blog post titled The Kimi K3 Moment argued that open-weight Kimi K3 is interchangeable with proprietary models for everyday coding work.