NVIDIA/OpenShell
Sandboxed runtime for fleets of autonomous agents, with kernel-level file, syscall and network policy enforcement.
Why it matters: Agents never see real credentials, and the timing could not be better.
Repository Watchnew on GitHub
kanaworksai-hue/audio-studio-mcp
Local background music and sound effects for Claude Code or any MCP client, via Stable Audio 3.
Why it matters: Runs on Apple Silicon or NVIDIA GPUs, with no cloud and no API key.
kukemc/bili-cli
A Bilibili command-line tool built for agents: login, watch, comment, message and upload, all with JSON output.
Why it matters: A clean pattern for wrapping a consumer platform so agents can drive it.
abdiraimov-otabek/dashboard-design-skill
An agent skill for honest dashboard UI: KPI tiles, charts, date pickers, empty states and a screenshot QA pass.
Why it matters: Distilled from 127 shipped dashboards, so it encodes taste, not just syntax.
gokayfem/video-to-blender
A coding-agent skill that turns a reference video or image into editable procedural Blender node systems.
Why it matters: Ships a tested node-group library rather than relying on the model to guess.
ethanplusai/astro-conversion-kit
Agent skills that convert a WordPress or other live site into static Astro with URL, SEO and pixel parity.
Why it matters: Timely for anyone patching PHP this week and wondering why they still run it.
The Orange Wiretop of Hacker News
-
Quad9 to exit France
A French court's blocking order sends the privacy DNS resolver out of the country on November 24.
-
Eight Bits of Trust: eMudhra's CA keys cracked in milliseconds
Scott Helme's explainer behind the front-page eMudhra story.
- Jailbreaking the Switch 2
- Today's PHP security release, and how we're strengthening security
- Every AI model has the same "controversial" take on humans
-
WorldTUI: I trained a 360M model to pretend to be a terminal
No shell and no filesystem: just a small model predicting the screen, running vim and tmux in WebGPU.
- The recently proposed o(n log n) integer multiplication algorithm [pdf]
- Qwen3-Coder writes less secure code when told it's for a US persona
- Using LLM for confidence scoring is a bad idea
- The compiler was supposed to end programming too