📦 anthropics/healthcare

Healthcare: The AI Plugin That Cites Its Sources

Healthcare is an AI plugin that turns messy medical paperwork into verified, actionable insights by actually showing its work. It tackles the nightmare of buried details in clinical notes and legal contracts by extracting specific facts and citing the exact line numbers where they appear. You can feed it a clinical note, and it will pull out billable codes or structured data, but the real magic is in its contract and document analysis. Instead of guessing, it scans the text, finds the relevant clauses, and stamps every claim with a verbatim quote and precise location.

🆔 @hackernewsgithubprojects

📦 orbien-org/orbien

Orbien: The 5MB Tool That Fixes Your Network

Orbien lets you securely expose any service on your local network to the internet, even if your router blocks incoming connections. It is a tiny, five-megabyte tool built in Rust that connects to a server and tunnels your traffic through it, supporting web apps, game servers, and SSH. What makes it special is how fast and light it runs, using modern protocols to keep things stable without needing a heavy Java environment. You get a simple desktop app to manage all your tunnels, so you can set up secure remote access in minutes. It is the perfect lightweight alternative for developers who need reliable, high-performance network tunneling without the bloat.

🆔 @hackernewsgithubprojects

📦 miaai-lab/qwen3.8-27b-dgx-spark-rtx-6000

Run a 1 Million Token AI Locally on Your Desk

This setup is the local server that finally makes running a massive AI model on a single desktop chip feel almost too easy. It takes a giant, compressed version of the Qwen3.8 model and serves it right from your own machine, meaning you can chat with an intelligence that can read an entire book in one go without sending your data anywhere. The coolest part is that it uses clever memory tricks to squeeze a one-million-token context onto standard hardware, so you get huge, smart, private answers without renting a massive cloud cluster. Just run three simple scripts to download, start, and use your very own high-power AI engine.

🆔 @hackernewsgithubprojects

📦 cvg/resplat

ReSplat: The Self-Correcting 3D Camera Model

ReSplat takes a handful of standard photos and turns them into a full three-dimensional scene by using a clever trick to fix its own mistakes. Instead of relying on complex math to correct errors, the model looks at the difference between the image it just rendered and the actual photo. It uses that visual gap to automatically adjust and improve the three-dimensional points it created, making the final scene much sharper and more accurate. This approach allows the system to build detailed environments with far fewer points than before, saving significant memory while keeping the quality high.

🆔 @hackernewsgithubprojects

📦 schapman1974/briskdb

BriskDB: Sharded SQLite Without The Fork

BriskDB takes your ordinary SQLite files and turns them into a single, sharded database that handles parallel writes without breaking a sweat. Instead of forking the engine, it splits your data across multiple standard SQLite files, allowing different parts of your system to write simultaneously without getting stuck in a traffic jam. The coolest part is that it speaks PostgreSQL and HTTP, so your existing tools just keep working while the backend scales. It even generates unique IDs across shards without needing a central bottleneck for every single insert.

📰 https://news.ycombinator.com/item?id=49312707

🆔 @hackernewsgithubprojects

📦 nagi-ovo/dsh-ads

dsh-ads: The Plugin That Fakes 2005 Web Ads

Dsh-ads is the plugin that turns a serious AI interface into a chaotic 2005 web portal. It injects fake popups, cartoon viruses, and spinning jackpot wheels into your chat. While the AI works, the screen pretends to pause for a 'rewarded' ad, only revealing the answer once the fake commercial ends. It even pulls real community plugins into these tacky banners. It is pure digital nostalgia, proving that even advanced tech can be made to feel delightfully broken and nostalgic.

🆔 @hackernewsgithubprojects

📦 teja2495/quick-search

Quick Search: Your Phone's New Brain

Quick Search turns your Android home screen into a universal command center that actually listens. Instead of digging through folders or hunting for apps, you type one thing and the app digs through everything on your device at once. It pulls up contacts, files, calendar events, notes, shortcuts, and even web results in a single, fast list. The coolest part? It handles fuzzy input, so if you mistype a name or forget the exact file location, it still finds the right thing. You can even jump straight into calls, texts, or web searches without leaving the app.

🆔 @hackernewsgithubprojects

📦 schmarta/claude-code-openai-server

The Free AI Bridge

claude-code-openai-server is the local bridge that makes your existing Claude Code login work inside any standard AI app. It wraps the command line tool in a web server that speaks the same language as OpenAI, so your chat apps can use that powerful model without paying for separate API tokens. The coolest part is that it keeps one single process alive for your entire conversation, meaning it remembers context and handles complex tool requests without the usual lag of starting over. Just point your favorite client at this local server and you get the full power of Claude for free.

🆔 @hackernewsgithubprojects

📦 zync-sh/zync

Zync: The All-In-One SSH Workspace

Connect to remote servers, manage files, and run terminals without juggling separate tools. Zync is an open-source desktop app that combines SSH connections, file transfers, and a built-in terminal into one fast, native interface. Instead of relying on a cloud server to store your data, it keeps your hosts, credentials, and settings local on your machine. You can even sync encrypted backups to your own personal cloud account. It’s a powerful, self-contained workspace for developers who want speed and privacy without subscriptions or vendor lock-in.

🆔 @hackernewsgithubprojects

📦 leonq8/comfyui-allinone-minimaxh3

One Node for All MiniMax H3 Video Creation

ComfyUI All In One MiniMax H3 is the node that finally makes video generation feel as simple as using a photo app. Instead of wiring a complex maze of connections or hunting through a dozen different plugins, you just pick one of several modes, drop in your prompt or reference images, and hit generate. This single tool handles everything from creating videos with native audio to animating still frames and even syncing mouth movements to your own soundtrack. It solves the headache of complex setup by hiding all the technical wiring behind a clean, user-friendly interface.

🆔 @hackernewsgithubprojects

📦 miaai-lab/qwen3.8-27b-sglang-dgx-spark

One Script for Qwen on DGX Spark

Run a massive AI model on a tiny computer with just two simple scripts. This project serves the Qwen 27 billion model on the NVIDIA DGX Spark, a compact desktop machine with massive memory. Instead of wrestling with complex server configurations, you simply copy a config file, run the start script, and you have a fully functional AI server ready to handle long documents and multiple users at once. It automatically handles the technical heavy lifting, like loading the right model version and setting up speed optimizations, so you can focus on using the tool rather than debugging the setup.

🆔 @hackernewsgithubprojects

📦 ucsandman/openclaw-hierarchical-memory-system

The Hierarchical Memory System

Give your AI agent a brain that actually scales. The OpenClaw Hierarchical Memory System replaces one bloated memory file with a smart index and detail folders, so your agent stops wasting tokens loading everything. Instead of reloading ten thousand characters of context every time, it starts with a tiny reference table and only digs into specific files when a conversation actually needs them. This cuts your session costs by roughly seventy percent while keeping critical details front and center. It is a simple, file-based upgrade that makes your agent faster, cheaper, and far more reliable.

🆔 @hackernewsgithubprojects

📦 ucsandman/openclaw-setup

OpenClaw Setup: The Complete AI Workspace

Give your AI assistant a permanent memory and a distinct personality using OpenClaw Setup. This repository provides a complete workspace for running an assistant that actually remembers who you are and what you're working on, solving the common problem of starting every chat from scratch. It includes a structured system of files that acts as a long-term brain, organizing notes about your projects and people so the assistant doesn't forget important details. The most interesting part is the reflection framework, which allows the assistant to review its own experiences over time and update its core personality files based on what it learns.

🆔 @hackernewsgithubprojects

📦 elayadesign/ai-design-skills

AI Design Skills

AI Design Skills is the design rulebook that makes your coding AI actually build beautiful landing pages. You know how AI code looks a bit generic and boring? This fixes that by giving your AI a strict set of visual and copywriting instructions. It stops the AI from using boring fonts or fake placeholder text, and instead makes it write real copy, pick smart layouts, and use smooth, modern animations. It is basically a cheat sheet for telling your AI exactly how to make a website look professional without you having to design it yourself.

🆔 @hackernewsgithubprojects

📦 jeremy-prt/bloub

How bloub recreates the x.ai avatar

Bloub is the web app that recreates the x.ai bot avatar by measuring its motion frame by frame. It captures sixteen distinct poses into a single black shape that morphs smoothly without any animation library. The developer didn’t guess the movements; he analyzed the original video to extract the exact angles and timings, so the eyes tilt and the body settles just like the real thing. It even lets you export the result as a video or image. This project proves that pixel-perfect accuracy is possible when you respect the original measurements instead of approximating them.

🆔 @hackernewsgithubprojects

📦 p2r3/ha.mr

Compressing Long Links Into Tiny Codes

The site ha.mr, which sounds like hammer, squeezes long internet addresses into tiny codes using pure math inside your browser. It works by stripping away the standard web address parts like https and www, then applying a compression method called Huffman coding to the remaining domain and text. Because everything happens locally in your browser without any server database, it is fast and private. When you click the short link, it instantly expands back to the original address so you can land exactly where you intended. You can also generate compact QR codes that use the most efficient character set for scanning.

📰 https://news.ycombinator.com/item?id=49326244

🆔 @hackernewsgithubprojects

📦 jitpass/jit

JIT Pass: Your Mac Secrets, Locked Down

Stop leaving your passwords in plain text files where any rogue script can grab them. JIT Pass, a free tool for your Mac, finds those exposed keys and hides them inside a secure vault protected by your fingerprint. The coolest part is how it works: your tools keep running exactly as they always have, but the secret only pops up in memory right when you need it. If a sketchy program tries to read your files while you are away, it just sees a fake placeholder. You unlock once with Touch ID, and the tool handles the rest without you changing a single command.

📰 https://news.ycombinator.com/item?id=49317546

🆔 @hackernewsgithubprojects

📦 shinshin86/live2d-add-motion-sample-web-ui

Live2D Add Motion Sample Web UI: Add Animations with JSON

Live2D Add Motion Sample Web UI is the development tool that lets you add new animations to characters without opening heavy editing software. Normally, making a 2D model blink or nod requires a complex rigging environment, but this project proves you can do it with simple text files. It uses a smart pipeline to turn basic code into valid motion data, ensuring the character actually moves correctly in your browser. You can drag, zoom, and test new poses instantly, making it perfect for developers who want to experiment with character behavior quickly. Skip the expensive setup and just focus on the creative side of animation today.

🆔 @hackernewsgithubprojects

📦 zyora-dev/zse

ZSE: The LLM Engine That Installs in 5 MB

ZSE is the inference engine that runs large language models without the heavy machinery usually required. While standard tools demand massive packages and minutes of setup, ZSE uses pure Python and a custom compiler to load models in seconds with a tiny memory footprint. This project proves you can build production-grade AI servers from scratch, eliminating complex dependencies while still supporting NVIDIA, AMD, and Apple hardware. It is a compelling look at how raw engineering can strip away bloat to deliver speed and simplicity for developers who need efficiency over convenience.

🆔 @hackernewsgithubprojects

📦 sylphai-inc/atskills

The End of Skill Installation

A new protocol called atskills treats agent skills like simple files you read on demand, rather than software you install. By using a file path as the address, the system loads instructions only when needed and leaves no permanent trace in your agent's memory. This solves the problem of clutter, where unused skills consume valuable space and create confusion. It turns a messy catalog into a clean, readable tree where you grab exactly what you need without any setup. The result is a cleaner workflow where your tools remain uncluttered and ready for action.

🆔 @hackernewsgithubprojects