📦 rayl15/openvision

openvision

Connect your Meta Ray-Ban smart glasses directly to powerful artificial intelligence running right on your iPhone. This neat app lets you swap between different brains for your glasses, including private models that run entirely offline on your device with absolutely zero API costs. The coolest part is the built-in face recognition, which lets you teach your glasses to recognize friends on the spot without ever sending a single photo or query to the cloud. It is like giving your wearable tech a completely local, private superpower. Grab your glasses and check out this open-source project to make them truly yours.

🆔 @hackernewsgithubprojects

📦 scstelz/security-investigator

Security Investigator

Run complete, multi-domain security investigations using simple conversational questions with security-investigator. This clever repository links GitHub Copilot, Model Context Protocol servers, and Microsoft Sentinel to automatically hunt for threats across your entire network. Instead of manually writing complex queries, you can just ask the assistant to investigate a suspicious user or check if an IP address is malicious. It automatically triggers specialized workflows, gathers data, runs verified search queries, and builds beautifully detailed reports complete with geolocation maps. It is like having a seasoned security analyst on standby, ready to audit everything from compromised credentials to misconfigured systems in seconds.

🆔 @hackernewsgithubprojects

📦 looptroop-ai/looptroop

LoopTroop AI

LoopTroop is the local orchestrator that finally stops AI coding agents from getting confused on complex tasks. Instead of relying on a single, messy chat session that gets bloated and drops in quality, this tool breaks big coding tickets down into bite-sized milestones called beads. It runs an LLM council where different models draft, vote on, and perfect a plan before writing any code. Then, an agent executes each milestone inside an isolated workspace, running automated testing loops that wipe the chat history and retry fresh if they fail. It is slow, precise, and perfect for leaving complex coding tasks to run safely overnight.

📰 https://news.ycombinator.com/item?id=48643477

🆔 @hackernewsgithubprojects

📦 majidmanzarpour/threejs-procedural-spider

threejs-procedural-spider

This web-based spider climbs walls, clings to ceilings, and leaps onto rocky terrain using pure math instead of pre-made animation clips or a physics engine. Built with Three.js, the project calculates every single movement on the fly using a self-stabilizing system. Its eight legs automatically find realistic footholds using formulas that solve the joints instantly, while its two-part body naturally sags and folds under its own weight when scaling vertical cliffs. Because the entire environment is built from equations rather than a standard 3D grid, the spider always knows the exact angle of the surface underneath it. It is an incredibly smooth demonstration of how math can breathe life into...

🆔 @hackernewsgithubprojects

📦 refactoringhq/portent

Portent

Most knowledge bases get ruined because we either build messy, endless folders or overcomplicate the setup from day one. Portent fixes this by introducing an open specification that acts as a universal blueprint for organizing personal and work databases. It strips away the chaos and replaces it with just eight default categories split between actionable work and reference memories, tied together by two basic relationships. By separating quick capture from deliberate organizing and long-term archiving, it keeps your system clean and completely portable across markdown files, note apps, and AI agents. It is a brilliant way to make your knowledge searchable and sensible without over-engineering your life.

🆔 @hackernewsgithubprojects

📦 prabal-rje/latentscore

LatentScore

LatentScore is the Python library that finally lets you generate custom ambient soundscapes using absolutely zero GPU power. Instead of burning through massive cloud compute bills or waiting on heavy neural networks, it uses a clever local embedding lookup that translates plain-text descriptions into synthesizer controls. You type a prompt like morning coffee shop or thunderstorm on a tin roof, and it instantly translates that vibe into custom audio parameters, outputting real sound directly to your speakers in just one line of code. It is an incredibly fast, lightweight way to generate procedural background music locally. Check it out and start scripting your own custom soundtrack today.

🆔 @hackernewsgithubprojects

📦 lzhgus/capso

Capso

Capso is the open-source screen capture app that finally brings premium features to your Mac for free. Instead of paying a steep subscription for tools like CleanShot X, you get a beautiful, native utility built with Swift and SwiftUI. It packs a floating control bar that lets you take screenshots, record smooth MP4s or quick GIFs, and even add a customizable webcam picture-in-picture. You can translate text on your screen instantly with built-in OCR, or spruce up your screenshots with gorgeous background gradients and drop shadows before sharing. Download Capso today to elevate your daily workflow without spending a dime.

🆔 @hackernewsgithubprojects

📦 jane-xiaoer/claude-skill-web-clone

Claude Skill Web Clone

Most AI tools generate fake, broken code when you ask them to copy a website, but this clever tool actually extracts the genuine, living source. Called claude-skill-web-clone, this methodology acts as a hands-on guide for your AI agents to faithfully mirror any web design, including highly interactive WebGL and 3D experiences, without relying on hallucinations. It uses browser probes and specialized scripts to capture real runtime assets and build a precise blueprint of the site's design DNA. It is the ultimate way to study advanced frontend techniques locally and safely swap in your own branding.

🆔 @hackernewsgithubprojects

📦 pku-yuangroup/ultrashape-1.0

UltraShape 1.0

You can now transform rough, low-resolution 3D models into highly detailed, watertight masterpieces using a clever two-stage diffusion process. The project is called ultrashape-1.0, and it fixes the biggest headache in 3D generation: flat details and messy, broken geometry. Instead of trying to build a perfect shape from scratch, this tool takes a coarse, basic 3D model generated from an image, locks onto its coordinates using spatial anchors, and focuses its computing power purely on carving out high-fidelity local details. It even includes a built-in data pipeline that automatically patches holes and thickens fragile structures. Download the pre-trained weights and give your rough 3D assets the ultimate geometric glow-up.

🆔 @hackernewsgithubprojects

📦 xt4d/gameblocks

GameBlocks

AI coding agents struggle to build 3D video games because translating spatial math into language tokens is incredibly fragile, often resulting in backward controls or broken physics. A clever open-source library called gameblocks solves this by giving AI assistants a robust set of ready-to-use coding blueprints. Instead of generating complex 3D math and physics interactions from scratch, agents can pull pre-made modules for vehicle steering, character movement, camera rigs, and terrain generation directly into their workspace. It acts as a bridge between language models and real-time game engines, keeping game states perfectly synchronized so developers can watch their AI-generated concepts come to life instantly.

🆔 @hackernewsgithubprojects

📦 jjenglert1/getting-started-with-claude-cowork

Getting Started with Claude Cowork

Get Claude's agentic desktop assistant running and doing real work for you today without touching a command line. The getting-started-with-claude-cowork repository is a hand-on companion guide packed with copy-and-paste templates, setup checklists, and practical prompt structures. The single coolest thing here is the viral workspace starter prompt, which instructs Claude to interview you step-by-step to automatically build custom memory, brand voice, and preference files directly in your folder. This creates a highly personalized, local AI brain that actually understands how you work. Clone this guide today to turn your desktop chat into an autonomous partner.

🆔 @hackernewsgithubprojects

📦 montevive/autocache

Autocache

Slash your Claude API costs by up to ninety percent without changing a single line of your code. If you build AI agents in tools like n8n or Flowise, you know how expensive it gets when your system prompts and tool definitions are sent over and over with every single turn. This clever tool runs as a local proxy that sits between your app and Anthropic. It automatically analyzes your requests, counts tokens on the fly, and injects smart cache-control breakpoints right where they belong. You just change your API base URL to point to this proxy, and you instantly get faster response times and massive savings. It even passes...

🆔 @hackernewsgithubprojects

📦 michaljach/opencode-browser

Opencode Browser

Opencode-browser is the browser automation plugin that directly connects your AI editor to Chrome or Edge for ultra-fast web tasks. Instead of wrestling with complex scraping scripts or manual testing setups, this tool integrates a specialized protocol to let your AI navigate sites, fill forms, and extract data in real time. What makes it special is its speed-focused optimization; it injects smart guidance directly into the AI's prompts to prevent redundant page reloads and eliminate artificial reconnect delays. It is the perfect helper for anyone wanting to automate web workflows instantly right from their development workspace.

🆔 @hackernewsgithubprojects

📦 paopaoandlingyia/prismcat

PrismCat

PrismCat is the self-hosted debugging proxy that finally reveals exactly what your apps are sending to language models behind the scenes. Think of it as a local black box recorder for your AI calls. Just change your base URL to route traffic through it, and you will instantly capture every raw request, response, and streaming chunk in a clean dashboard. If an agent goes rogue or a prompt breaks, you do not have to rerun your script. You can tweak parameters and replay the exact API request directly in your browser. It runs locally via SQLite, keeping your API keys and data completely private.

📰 https://news.ycombinator.com/item?id=48279360

🆔 @hackernewsgithubprojects

📦 qasimwani/spatial-reasoning

Spatial Reasoning

Pinpoint highly specific, hard-to-find objects in any image just by describing them in plain English. This Python package, called spatial-reasoning, lets you run advanced object detection without needing a custom-trained model. Instead of relying on rigid, pre-set categories, it pairs powerful vision models with smart, grid-based reasoning tools. You can ask it to find the farthest scooter on a busy street or a tiny, obscure detail in a photo, and it logically breaks down the image to locate it. It is incredibly easy to set up and brings highly flexible visual search right to your local code.

🆔 @hackernewsgithubprojects

📦 trx7111/comfyui-trixloader

comfyui-trixloader

You can now edit, crop, and mask your images directly inside ComfyUI without ever jumping back and forth to Photoshop. The comfyui-trixloader project brings a complete image editing suite right into your node canvas. It features a Lightroom-style workspace with adjustment sliders, color curves, and layer controls. Even better, it packs a built-in AI mask editor that lets you isolate objects instantly using advanced background removal and text-prompted masking. Whether you need to fix colors, pad edges, or generate complex masks, this clever tool lets you handle your entire image prep pipeline right inside your workflow.

🆔 @hackernewsgithubprojects

📦 vg-gui-tasker/vg-gui-tasker

vg-gui-tasker

Teach digital agents to navigate software simply by watching video tutorials. While AI models can answer basic questions about a video, they struggle to translate those visual lessons into actual step-by-step actions on a computer screen. This project closes that gap with a smart search algorithm called TASKER that acts like a human eye, instantly pulling out only the most important video frames needed to complete a task. By filtering out background noise and focusing on key moments, it helps GUI agents click, scroll, and type their way through complex online workflows with incredible accuracy and efficiency.

🆔 @hackernewsgithubprojects

📦 juntaojianggavin/april-medseg

April MedSeg

Mix and match any neural network component for medical image analysis with just a single line of configuration. Setting up medical image segmentation usually means wrestling with rigid, hardcoded architectures, but april-medseg changes the game by turning the entire process into a modular playground. You can instantly swap between over a hundred encoders, dozens of decoders, and specialized skip connections using simple configuration files. Whether you are experimenting with classic architectures, cutting-edge foundation models, or advanced training methods like semi-supervised learning, this tool lets you assemble the perfect pipeline without writing custom glue code. Grab your coffee and start building smarter medical AI systems today.

🆔 @hackernewsgithubprojects

📦 danyalrehman/autobg

autobg

Ditching complex physics simulators to predict how proteins wiggle and fold just got a massive upgrade because we can now generate molecular structures using the same core math that powers modern chat bots. A new project called autobg completely throws out the traditional, slow, math-heavy simulation flows. Instead, it breaks down the exact three-dimensional coordinates of a molecule's atoms and predicts them step-by-step like words in a sentence using a custom transformer. It is incredibly fast, avoids the typical geometric errors that break other models, and its pre-trained model can even predict structures for entirely unseen peptides on the first try.

🆔 @hackernewsgithubprojects

📦 yao-ustc/adapt

ADAPT

adapt is the tool that finally stops vision-language models from hallucinating objects that are not actually in your images. When a visual model writes a long description, its focus naturally drifts away from the picture and starts making things up based on text patterns. This project fixes that by tracking the model's early, highly accurate visual focus to build a stable anchor. If the model's attention starts to wander during writing, the system steps in and gently steers its focus back to the right parts of the image. Give it a spin to build incredibly accurate and visually grounded AI applications.

🆔 @hackernewsgithubprojects