DevHunt Daily
Monday, November 9, 2026
- 1AIHacker News
Llama-modchip enables 284B DeepSeek inference on 64 GB RAM without GPU
The llama-modchip project adds CPU‑first support for DeepSeek's 284‑billion‑parameter model, allowing inference on machines with 16–64 GB RAM and no GPU. Developers can run large Mixture‑of‑Experts models locally using only system memory.
- 2Open sourceHacker News
BenchUp launches open-source 3D modeling app for woodworking built with Rust and WebGPU
BenchUp is a free, open-source desktop and browser‑based 3D design tool aimed at woodworkers and furniture makers. It runs on Windows, macOS, Linux and in browsers with WebGPU/WebGL2, offering sketching, push‑pull modeling, exact measurements and import/export of SketchUp and OBJ files. Developers may be interested in its Rust codebase and WebGPU implementation.
- 3Open sourceHacker News
Dux releases version 2.1 with new system monitor and UI enhancements
The open‑source Wayland desktop shell Dux announced version 2.1, adding a system monitor, video wallpapers, an updated media player, and various UI improvements. The update also expands support for multiple Wayland compositors and introduces a modular shell with theming and widget customization, which may interest developers building Linux desktop environments.
- 4AIGitHub Trending
Claude Quickstarts provide ready‑to‑run examples for building Claude API applications
Anthropic released a collection of starter projects that demonstrate how to use the Claude API for tasks such as customer support, financial analysis, computer and browser automation, and autonomous coding. Developers can clone these examples to accelerate building and customizing Claude‑powered applications.
github.com32 stars today - 5CloudGitHub Trending
NVIDIA releases NVKE preview for managing GPU Kubernetes clusters
NVIDIA Kubernetes Engine (NVKE) provides a management plane to create and operate Kubernetes clusters on NVIDIA GPU hardware. It runs in preview, supports local evaluation via Docker and Colima, and requires Go and Make for building from source, which is relevant for developers building GPU‑accelerated workloads.
github.com345 stars today - 6AIGitHub Trending
billion-context adds context-compression plugin for long sessions with token savings
billion-context provides a plugin that compresses context, enabling month‑long sessions with billions of tokens while using small 100K windows and reducing token usage by up to five times. Developers working with AI models can keep longer conversations efficiently.
github.com460 stars today - 7Dev toolsLobsters
systemd adds optional Rust support for new components
systemd merged Rust support, allowing new leaf binaries to be written in Rust while keeping core libraries in C. The integration follows strict guidelines on API uniformity, build system, and dependency control, ensuring existing behavior remains unchanged.
- 8Dev toolsLobsters
nspawn 3.0 adds Docker-compatible Engine API for systemd-nspawn containers
nspawn 3.0 provides a Docker Engine API socket on the host, allowing Docker CLI tools like GitLab Runner to manage systemd-nspawn containers as if they were Docker containers. Developers can use standard Docker commands against /run/nspawn/api.sock, enabling compatibility with existing Docker workflows.
- 9Open sourceLobsters
libghostty 1.0 released with stable C ABI
The libghostty-vt C library for terminal emulation reached version 1.0, marking a stable API and ABI. It is now used in JetBrains IDEs, Neovim, KDE, Vercel Turborepo and many other projects, enabling reliable packaging and dynamic linking for developers.
- 10WebLobsters
Meridian browser for macOS released, built on Rust, GPUI and Servo
Meridian is a macOS web browser written in Rust that uses the GPUI UI framework and the Servo engine. It offers configurable fonts, keybindings, vertical/horizontal tabs, reader mode and other features, and can be installed via Nix. Developers may be interested in its Rust codebase and the gpui-servo integration.
- 11AIHacker News
Agent memory formats lack standard, call for OpenTelemetry-like agreement
The blog notes that various AI agents store memories in incompatible formats—from markdown files to JSON stores—preventing tools from sharing data. It argues for a community‑wide specification similar to OpenTelemetry to unify data models, conventions, and protocols, which would simplify development of agent‑based systems.
- 12SecurityHacker News
PacketHound adds ask‑first outbound firewall for Windows
PacketHound is a free Windows utility that turns the built‑in firewall into an outbound "ask me first" filter. It prompts the first time an app tries to connect to the internet, lets you allow or block, and creates persistent firewall rules. Developers may use it to monitor and control network traffic from their tools.
