📦 urbanground/urbanground

UrbanGround: Testing AI in Real Cities

UrbanGround is the realistic city simulator that actually tests if AI can walk down a real street. Most AI agents play in video games, but UrbanGround takes the entire three-dimensional layout of Hong Kong and turns it into a playable world where a language model has to navigate using only what it sees. You can literally watch the AI figure out how to climb stairs or find a coffee shop, and the project includes seven hundred specific challenges to see if it truly understands the space or just gets lucky.

🆔 @hackernewsgithubprojects

📦 irajisamurai/2d-motion-interface

2D Motion Interface: Captioning Video Without 3D

2D Motion Interface is the lightweight tool that lets AI describe human movement from flat video without needing expensive 3D tracking. Most motion models require deep, volumetric data that is hard to capture in the real world, but this project solves that by training a small bridge module. It teaches a standard 2D pose estimator to speak the same language as a 3D model, allowing it to generate accurate text captions from a simple phone camera. The surprising part is that this flat, 2D input actually outperforms traditional 3D estimates on real-world video, proving you can get high-quality results with far less computational power.

🆔 @hackernewsgithubprojects

📦 rarf/hermes-quota-plugin

Hermes Quota: Stop Guessing Your AI Limits

Hermes Quota is the desktop plugin that actually shows you how much AI usage you have left. It solves the mystery of hidden rate limits by pulling live data from providers like Anthropic and OpenAI directly into your sidebar and status bar. You get a clear progress bar for every account, so you know exactly when your credits reset without opening ten different tabs. The best part is it works quietly in the background, keeping your workflow smooth while it tracks your usage. No more guessing when you run out of tokens. It just works, giving you total control over your AI budget in one clean view.

🆔 @hackernewsgithubprojects

📦 links-ads/geoid-flood

GEOID-Flood: The Global Flood Dataset

GEOID-Flood is the satellite dataset that finally maps real-world floods across the globe with pinpoint accuracy. It pulls from two hundred nineteen actual disaster events spanning sixty five countries, turning them into thousands of high resolution images. The cool part is that it doesn't just show where the water is. It pairs satellite photos with terrain maps to teach computers the difference between a river that is always there and a river that has burst its banks. This gives researchers a huge, clean library to build systems that can spot dangerous flooding before it gets worse.

🆔 @hackernewsgithubprojects

📦 eximius-labs/fusion-embedding

Fusion Embedding: One Space for Seven Senses

A single model can now match a dog howling to the exact photo of that dog, without ever seeing a pair of audio and image during training. Fusion Embedding is an open-weight system that maps text, images, video, audio, thermal, motion, and touch into one shared vector space. It works by freezing a powerful base model and adding small, pluggable sensor heads, which means you can add new types of data without retraining the entire system. This is incredibly useful for developers who need to search across different types of media simultaneously, such as finding a video clip by describing its sound.

🆔 @hackernewsgithubprojects

📦 pollen-robotics/microduck_rl

Microduck RL: Teaching a Tiny Robot to Do Tricks

This tiny, two-legged robot learns to do forward rolls and kick balls by practicing in a physics simulator that treats gear play and motor friction as real. The Microduck RL project provides a complete toolkit for training an eighty-gram bipedal robot to walk, stand up, and perform complex maneuvers. It solves the problem of the gap between simulation and reality by modeling the specific quirks of the robot's small servos, including the physical slack in its joints. By training policies that account for these imperfections, the robot can successfully execute learned tricks on actual hardware.

🆔 @hackernewsgithubprojects

📦 styles01/sparkrun-recipes

SparkRun Recipes: Run 180B AI on One Computer

SparkRun Recipes is the collection of tested guides that lets you run massive artificial intelligence models on a single desktop machine. The standout feature is a recipe for a 180B parameter model that fits on one small computer by storing a huge lookup table on the hard drive instead of in memory. This clever trick allows the system to serve incredibly complex answers without needing a massive server rack. The repository includes detailed instructions for various models, showing exactly how to set them up for real-world tasks. It turns a simple home setup into a powerful local brain, giving developers a clear path to use big AI without the cloud.

🆔 @hackernewsgithubprojects

📦 linklyai/best-skills

Best Skills: The Cross-Ecosystem Agent Skill Leaderboard

Best Skills gives you a single, daily-updated leaderboard for AI agent skills by merging install data from major platforms like skills dot sh, ClawHub, and Tencent SkillHub. While other charts only show one platform's view, this project combines global downloads with social buzz from X, Hacker News, and Bluesky to reveal which skills are actually gaining traction everywhere. It tracks over ten thousand skills and publishes nine distinct rankings, from top installs to trending growth, all in open CSV format. Because the raw numbers for each platform are preserved side by side, you can verify the data or build your own analysis.

📰 https://news.ycombinator.com/item?id=49417363

🆔 @hackernewsgithubprojects

📦 elyosh/opentie

OpenTIE: The Best of Both Worlds

Pilot through the galaxy by bringing the classic Star Wars TIE Fighter fully to life on your modern computer. This open-source project runs the original game files on Windows, macOS, and Linux, solving the problem of outdated hardware and bringing the 1990s hit into the twenty-first century. The coolest part is the freedom it offers: you can mix the immersive 1998 graphics and flight physics with the reactive, adaptive music from the 1995 release. You don't have to choose between pretty visuals or a great soundtrack; you get the best of both worlds in one smooth package.

📰 https://news.ycombinator.com/item?id=49471965

🆔 @hackernewsgithubprojects

📦 icastanon/fintrack-platform

FinTrack: The Event-Driven Finance Platform

FinTrack is a personal finance platform that treats every financial change as an event, allowing the system to automatically categorize spending and check your budget without waiting for your manual input. Built with Java and Spring Boot, it separates the user interface from the heavy lifting, using a message queue to handle transaction imports and balance updates in the background. This setup means if you upload a large file of transactions, the system processes it in chunks that can recover from errors without losing your work. It is a practical, open source tool that demonstrates how modern cloud infrastructure can make personal budgeting both reliable and transparent.

🆔 @hackernewsgithubprojects

📦 microsoft/hve-core

HVE Core: Your AI Team

HVE Core is the project kit that finally gives your AI assistant the specific job roles it needs to actually ship work. Microsoft built this open source collection to solve the problem of generic AI advice by providing pre written instructions, prompts, and specialized agents for everything from security reviews to accessibility checks and sprint planning. You can drop these components into your existing GitHub Copilot setup to instantly upgrade your workflow. It turns your coding assistant from a helpful buddy into a disciplined team that follows your company standards without you having to re teach it every day.

🆔 @hackernewsgithubprojects

📦 uynitsuj/warp-rm

Warp RM: Scoring Robot Video Progress

Turns raw robot videos into a precise measure of task progress, solving the problem of figuring out which demonstration clips are actually useful for training. Traditional methods often treat all footage as equal, but this tool calculates a signed velocity for every single frame, distinguishing between fast forward movement, stalling, and even backward regression. By using time-warping augmentations, it learns to predict this pace without needing expensive manual labels. This dense signal acts as a filter, allowing downstream policies to focus only on the most effective parts of the data, which dramatically improves how robots learn complex tasks like folding clothes or sorting objects.

🆔 @hackernewsgithubprojects

📦 subev/libratory

Libratory turns your PDFs into audiobooks

Libratory turns the messy PDFs on your hard drive into polished, chapter-marked audiobooks that run entirely on your own computer. The project acts as a digital workbench where you can clean up scanned text, translate chapters, and even chat with your entire library using local AI models. It synthesizes audio using local voices so your data never leaves your home. The key takeaway is that it transforms a static file into a living library, giving you full control over your digital reading experience without relying on cloud services.

🆔 @hackernewsgithubprojects

📦 plannotator/herdr-annotate

Annotate Your Terminal Output for AI

Pin comments to your terminal output so you can hand perfect context to your AI agents. Herdr Annotate solves the headache of losing track of which code snippet or error log you meant to reference when chatting with a bot. It lets you select text in your terminal, add a quick note like 'this is the bug' or 'use this config,' and save it. The coolest part? It turns all those scattered notes into clean, organized Markdown that you can copy with one click. This means you stop wasting tokens explaining where things are. You just paste your context, and the agent already knows what you are looking at.

📰 https://news.ycombinator.com/item?id=49457376

🆔 @hackernewsgithubprojects

📦 browser-use/benchmark

Benchmark: Testing Browser AIs

Browser-use just released a massive, open-source test suite designed to measure how well AI agents actually handle real web tasks. The core of this repository is a set of standardized challenges that force AI systems to navigate live websites, interact with forms, and solve puzzles under strict evaluation rubrics. Instead of guessing which framework is better, this project lets you run the exact same tasks across different tools to see who genuinely succeeds. It removes the guesswork from choosing AI browser agents by providing clear, reproducible scores. You get a fair, transparent way to compare performance and cost, so you can pick the right tool with confidence.

🆔 @hackernewsgithubprojects

📦 supermansundies/comfyui-klein-edit-composite

The AI Edit Fix

Klein Edit Composite is the ComfyUI node that actually makes AI edits look real. You know that annoying color shift when you use an AI tool to tweak a photo? This fixes it. It figures out exactly what you changed, blends only that part back in, and keeps the original background perfectly intact. It even matches the lighting so the new stuff doesn't look pasted on. No more weird halos or washed-out colors. Just clean, natural edits that look like you took the shot that way. If you are working with AI images, this is the fix you have been looking for. Your edits will finally look professional.

🆔 @hackernewsgithubprojects

📦 raress96/dolby-atmos-encoder

Dolby Atmos Encoder

A Rust tool called Dolby Atmos Encoder successfully converts high-end cinema audio into a stream that passes every available software verification test, yet it refuses to work on certified hardware. This project bridges the gap between lossless movie files and consumer home theater systems by translating complex object-based audio into a compressed format that standard receivers can understand. It parses the raw data, renders the sound, and injects the necessary metadata into the stream. However, a deep dive reveals that Dolby locks the final step behind a proprietary cryptographic signature that open-source tools cannot generate. The result is a perfect digital artifact that software accepts, but hardware ignores.

📰 https://news.ycombinator.com/item?id=48498705

🆔 @hackernewsgithubprojects

📦 streamcoreai/streamcore-server

StreamCore: The Voice Server That Lets You Cut In Mid-Sentence

StreamCore is the realtime voice server that finally lets your AI handle interruptions without stuttering. It takes the messy, latency-heavy stuff out of voice interactions so you can focus purely on the brain of your agent. The magic is its barge-in capability, which means if a user talks over the AI, the system instantly ducks the audio and cancels the current response to listen back. It handles the tricky network connections and streams speech in real time, so you don't have to wait for a full sentence to finish generating before the audio starts playing.

📰 https://news.ycombinator.com/item?id=49479199

🆔 @hackernewsgithubprojects

📦 crmne/fastpotify

Fastpotify: The Spotify Client That Actually Feels Native

Fastpotify lets you ditch the bloated web browser and run your entire Spotify library in a lightweight, native desktop app that starts in under a second. Built with Rust, it plays music locally at high quality while acting as a full Spotify Connect device, so you can stream to your computer or control speakers on your network from one simple window. It handles playlists, search, and album art without lagging your system, giving you the familiar Spotify layout with the speed of a true native tool. If you hate waiting for a music app to load, this is the smooth, fast alternative you have been waiting for.

📰 https://news.ycombinator.com/item?id=49479389

🆔 @hackernewsgithubprojects

📦 merit-systems/openinstinct

OpenInstinct: The iMessage Assistant That Books Your Tickets

OpenInstinct is the open-source iMessage assistant that actually uses a real browser to do your chores. Instead of just giving you links to movie sites, it logs in, navigates the pages, and books your movie tickets or orders groceries right from your text thread. The best part? It runs entirely on your own infrastructure, meaning your sensitive passwords and credit card details stay encrypted and never touch third-party servers. It’s a personal digital butler that handles the annoying web work for you, so you can just text it and let it get things done while keeping total privacy.

📰 https://news.ycombinator.com/item?id=49479314

🆔 @hackernewsgithubprojects