📦 uwmilab/lcvn

LCVN: The Robot That Imagines The Walk

LCVN is the visual navigation system that actually imagines the future. Most robots get lost because they rely on constant live feedback to stay on track, but this model changes the game entirely. It takes a single photo and a simple human instruction, then predicts the entire rest of the journey without looking back. It’s like a robot with a mental map that can see through time, letting it plan a path before it even takes a single step. This work was even picked as an oral presentation at NeurIPS, proving the idea really works.

🆔 @hackernewsgithubprojects

📦 arcships/rutis

rutis

rutis is the plugin runtime that lets your software write its own upgrades without crashing. Instead of hard-coding features, agents can generate new plugins in languages like Rust, Python, or Go, and hot-swap them while the main program keeps running. The magic happens because the system tracks exactly what depends on what, so when a new version loads, only the affected parts restart safely. It is a self-healing infrastructure that keeps everything running while it evolves. If you want your apps to improve themselves on the fly, this is how you do it.

🆔 @hackernewsgithubprojects

📦 sparkgeo/geo-mcp-servers

Geo MCP Servers

Give your AI assistant direct access to the planet’s data. The Geo MCP Servers project connects language models to over a hundred real-world mapping tools, including satellite imagery, live weather feeds, and city routing engines. Instead of guessing, your AI can pull actual coordinates and spatial data to answer complex questions. This curated list tracks which servers are active and which are paid, saving you hours of digging through documentation. It turns a text chat window into a powerful geographic workbench. If you build with location data, this is the single source of truth you need to start.

🆔 @hackernewsgithubprojects

📦 moritzwilksch/diffle

Review Git Diffs Like a Hacker with Diffle

Review any Git diff directly in your browser while keeping your hands on the keyboard. Diffle is a local tool that turns your terminal into a powerful code review interface, letting you jump between files and hunks with simple keys. The standout feature is its deep integration with language servers, so you can jump straight to function definitions or check variable types right from the diff view without switching to an editor. It also connects to GitHub, allowing you to send your comments directly as a pending review.

📰 https://news.ycombinator.com/item?id=50006930

🆔 @hackernewsgithubprojects

📦 wieslawsoltes/vb6

VB6 in your browser

Someone rebuilt the entire Visual Basic six studio to run entirely inside your web browser. It is not just a simulator; it is a full development environment where you can drag and drop buttons, write code, and debug apps using modern web graphics. The coolest part is that you can export your finished project as a single, standalone HTML file that runs anywhere without installing anything. This is perfect for preserving legacy tools or building quick demos without heavy setup. No server, no database, just code. If you have old Visual Basic projects sitting on a hard drive, this is your ticket to reviving them in seconds.

🆔 @hackernewsgithubprojects

📦 shaongitbd/doory

Doory: The Open-Source Canva Alternative

Doory is the open source design editor that finally lets you own your creative tools completely. It runs entirely in your browser, so you can create posters, social posts, or presentations without paying a dime for a premium plan or sharing your data with a corporation. The standout feature is its one click background removal, which uses an artificial intelligence model that runs directly on your own computer rather than sending your images to a cloud server for processing. This keeps your work private and fast while allowing you to export designs as high quality images or even animated videos.

📰 https://news.ycombinator.com/item?id=50049872

🆔 @hackernewsgithubprojects

📦 buffalo2024/rulora

Rulora: The Safety Net for AI Agents

Rulora forces AI agents to stop when they start looping on the same task without making progress. It acts as a control layer for artificial intelligence, ensuring that outputs are validated and decisions are locked in before being delivered to users. The project separates the creative work of the model from the rigid checks of the program. This prevents agents from drifting off course or accepting bad data. You can drop these controls into existing systems to make AI behavior predictable. It is about making sure your software knows when to say no. The result is a safer and more reliable user experience.

🆔 @hackernewsgithubprojects

📦 jhaizhou-ops/d2r-mobile

Play Diablo 2 on Your iPhone

Run the full Nintendo Switch version of Diablo 2 Resurrected directly on your iPhone without any emulation or cloud streaming. This open source project rewrites just a handful of processor instructions so the original game code runs natively on Apple chips, bypassing the need for a slow emulator. You get a native sixty frame per second experience with a completely redesigned touch control scheme that feels natural on a phone. The repository includes new endgame content like the Greater Rift and a boss rush mode that you cannot find on the original console.

🆔 @hackernewsgithubprojects

📦 zhuyansen/new-term-radar

Spot AI Trends Days Before Google Trends

Catch brand new AI models and tools the moment they appear, skipping the wait for Google Trends to catch up. New Term Radar scans X posts, Hugging Face trending lists, and explosive GitHub repos to spot names that are about to blow up. It filters out common words to highlight genuine new products, giving you a head start on what people will search for next. It turns raw data into a daily report of emerging names before they hit major keyword tools. This is perfect for developers and marketers who need to know about the next big thing before it becomes mainstream.

🆔 @hackernewsgithubprojects

📦 samsaffron/term-llm

Term LLM: The AI Terminal All-In-One

Run your entire AI workflow straight from your terminal. Term LLM acts as a Swiss Army knife for your command line, letting you chat with models, edit files, and execute complex jobs without ever leaving the shell. Instead of juggling different tools for text and code, you get a single interface that connects to various AI providers. The coolest part is the live voice feature, which lets you talk to your machine and get real-time audio responses. It feels like having a helpful assistant who lives directly in your terminal window.

🆔 @hackernewsgithubprojects

📦 bennettxai/founderos

Founder OS: The AI Department for Your Business

Founder OS is the business command center that treats your company like a set of AI assisted departments. If you run a one person business, this tool turns the chaos of your tabs into one live dashboard where named AI agents handle specific jobs like finances, social growth, and client funnels. The most interesting part is its knowledge graph, which acts as a shared brain for these agents to ensure they learn from your data without getting confused. It is a powerful way to see your entire operation at a glance. The takeaway is simple: you can run a complex business with AI help from a single, clear screen.

🆔 @hackernewsgithubprojects

📦 hexuxiu/m365-copilot2api

Turn Copilot Into an API

Take a Microsoft 365 Copilot subscription and turn it into a standard API that your favorite coding tools already speak. This little Go app acts as a bridge, translating the private, messy protocol hiding behind your Copilot subscription into clean, standard endpoints that OpenAI and Anthropic clients expect. Suddenly, your editor or CLI tool can talk directly to the models behind Copilot, using the exact same format they use for everything else. No more custom hacks, no more format mismatches. It just works, letting you leverage the power of your existing subscription in a way that is actually usable in your daily workflow.

🆔 @hackernewsgithubprojects

📦 deeplethe/solid-glass

Solid Glass: The Industrial AI Theme

Solid Glass turns standard web forms into heavy industrial control panels. It takes a common design library and wraps it in thick smoky glass that feels like solid surface material rather than a thin overlay. This is built specifically for operator consoles and artificial intelligence dashboards where visual weight matters more than playful color. The creator DeepLethe focused on making every button and input look substantial and clear at the same time. It solves the problem of digital interfaces feeling too light for serious technical work.

🆔 @hackernewsgithubprojects

📦 h0m3us3r/w1700k-sdr

W1700K SDR: Turning a Wi-Fi Router Into a Radio

A standard Wi-Fi router can become a high-speed spectrum analyzer that sees the entire airwaves at once, without any proprietary vendor software. The project W1700K SDR rewrites the internal radio firmware from scratch to capture raw data at speeds up to 320 million samples per second. It streams this live to your computer, where a simple web page displays the spectrum and waterfall in real time. This is fascinating because the developers reverse-engineered the hardware to bypass the manufacturer’s locked-down code, creating a transparent receiver you can fully control. It turns a cheap commodity device into a powerful research tool for anyone interested in radio signals.

📰 https://news.ycombinator.com/item?id=50055593

🆔 @hackernewsgithubprojects

📦 sakanaai/mass

How a Model Learns From Its Own Team

MASS lets an artificial intelligence get better by creating copies of itself to work together on a problem. Instead of one model guessing the answer, a group of identical models collaborates to find the best solution. The system then studies the successful steps from that team to teach the original model new skills. This creates a loop where the model improves its own decision-making based on its past performance. It is a fascinating look at how systems can learn from their own actions without needing human labels. This approach offers a clear path toward self-improving software.

🆔 @hackernewsgithubprojects

📦 mistvvk/xestrata

Run 125 Billion Parameters on One Card

XeStrata is the inference engine that runs a massive one hundred twenty five billion parameter AI model on a single graphics card without requiring proprietary Nvidia drivers. It allows you to host complex language models locally using only one graphics processor from Intel, AMD, or Nvidia on standard Linux hardware. This project removes the need for expensive multi card server clusters by efficiently splitting the model across available memory and using native hardware acceleration. You no longer need a data center to experiment with state of the art artificial intelligence. This brings powerful local computing to your personal machine.

🆔 @hackernewsgithubprojects

📦 snuri00/ps2-web-recomp

Play PS2 Games in Your Browser Without an Emulator

PlayStation 2 games run directly in a web browser without using a traditional emulator. The project, called PS2 Web Recomp, takes the raw machine code from a game disc and translates it into C++ before compiling it into WebAssembly. This method creates a native version of the game that runs on modern hardware, rather than simulating the old console step by step. It currently supports titles like Gran Turismo 4 and God of War II, allowing them to play at full speed with graphics and sound. You bring your own disc image, and the project turns it into a playable webpage on your own computer.

📰 https://news.ycombinator.com/item?id=50052251

🆔 @hackernewsgithubprojects

📦 ayzzz123/mcd-coupon-strategist

mcd-coupon-strategist

Stop leaving money on the table. This tool calculates the cheapest way to order McDonald's by checking real-time prices and your specific coupons. It finds the best deal, but its killer feature is catching 'coupon traps.' You might think using a discount code saves money, but sometimes adding items just to use that code actually costs more. This project detects those traps and warns you. It even checks your past orders to show how much you waste when coupons expire. It turns vague saving habits into clear, actionable data. Use it to see your true spending and actually keep more cash in your pocket.

🆔 @hackernewsgithubprojects

📦 cordwainersmith/claudoscope

Claudoscope: See What Your AI Agent Is Actually Doing

Claudoscope turns the invisible activity of your AI coding agent into a clear, real-time dashboard on your Mac. It reads the local files where your agent stores its work and shows you exactly what it is doing right now, including whether it is stuck waiting for your approval. The most surprising part is that it watches for accidental leaks, alerting you if your agent accidentally pastes a password or private key into a chat. Because everything stays on your computer, you get instant visibility into costs and security risks without sending your data anywhere.

🆔 @hackernewsgithubprojects

📦 trhhubsd/artexx

Artexx: The AI Hacker That Won Baidu's Security Challenge

Artexx is an autonomous artificial intelligence system that won Baidu's agent plus attack and defense competition. This project functions as a digital penetration tester that independently navigates complex web applications to map out every hidden data endpoint. It solves the massive problem of manual security auditing by automating the discovery of unprotected interfaces in modern software stacks. This is useful because it gives developers a proactive way to find vulnerabilities before malicious actors do. The system uses intelligent agents to mimic human interaction with digital systems. It identifies weak points in code logic and data flow. You get a comprehensive map of your application's security posture.

🆔 @hackernewsgithubprojects

📦 morluto/awesome-reverse-engineering

Awesome Reverse Engineering: Your New Toolkit

You can literally look inside a computer program to see exactly how it thinks, and there is a dedicated map for doing it. The repository called Awesome Reverse Engineering is a curated directory that tells you which tool to pick for a specific job, like checking if a file is safe or reading the code inside an app. Instead of guessing, you just match your task to a section in the list. It is the perfect starting point for anyone curious about how software actually works under the hood, giving you a clear path without the guesswork.

🆔 @hackernewsgithubprojects