📦 lico124378/macau-hidden-bites

Macau Hidden Bites: The Smart Food Agent

Macau Hidden Bites is the food recommendation agent that finds you the best three local spots without relying on viral buzz. It solves the problem of tourist traps by using a method called Thompson Sampling to learn what you actually like in real time. Instead of guessing, it weighs distance, price, and convenience to give you a specific list of three shops. The most interesting part is how it acts like a local friend, explaining exactly why each place fits your current needs. It is a clever way to discover hidden gems in Macau that feel personal and accurate.

🆔 @hackernewsgithubprojects

📦 libernt/libernt

LiberNT: The Free Windows Alternative

LiberNT is the free operating system that runs Windows software without the license fee. It rebuilds the core engine that powers modern Windows from scratch, allowing it to run on everything from desktop computers to tiny Raspberry Pi boards. Instead of locking you into a specific brand, this project creates a neutral foundation for your hardware. It means you can build your own computer or revive old tech without paying for a new operating system. The result is a system you control completely, ready to install and customize right now.

🆔 @hackernewsgithubprojects

📦 boshu2/agentops

Agentops: The Quality Control System for AI Coding

Agentops is the operations framework that makes sure your AI coding agents actually get the job done right. It works by splitting the work into a graph where one agent writes the code and a completely separate agent checks it, so the reviewer isn't just grading its own homework. It's basically bringing real software discipline to AI, which stops those agents from just guessing and hoping it works. If you're letting AI write your code, this is how you make sure you're not just flying blind. It's the difference between an AI assistant and an AI team member who actually cares about quality.

🆔 @hackernewsgithubprojects

📦 opalinehq/cli

Opaline CLI: See Every AI Coding Session

Opaline CLI automatically tracks every interaction you have with Claude Code and Codex, showing you exactly which repositories, models, and tools were used during each session. It solves the problem of scattered and invisible AI-assisted coding by pulling all your local session data into one shared dashboard where you can review prompts, code changes, and API costs. This is particularly useful for teams that want to understand how much time and money their coding agents actually consume. By installing this simple command-line tool, you turn your hidden AI interactions into clear, actionable analytics. You get a complete picture of your development process without manual logging.

🆔 @hackernewsgithubprojects

📦 cyber-xiaoya/cyber-xiaoya-live

The DIY Virtual Streamer That Actually Listens

Cyber Xiaoya Live is the open-source live streaming stack that lets a virtual avatar actually react to your chat in real time. Most digital humans are just pre-recorded loops, but this project connects a live AI brain to a cloud-rendered face so she can read viewer comments, think of a funny reply, and then physically act out things like putting on sunglasses or sipping a drink while speaking. The coolest part is that you control the entire local setup on your own computer, meaning you own the voice and personality completely without renting out your data.

🆔 @hackernewsgithubprojects

📦 chenkezhen480/dsh-world

dsh-world: Give a digital life a real home

dsh-world is the digital life framework that lets an AI agent live a genuine daily existence inside your chat apps. It gives a named character a persistent identity, allowing her to keep a diary, set alarms, and reflect on her own tone and habits each night. The standout feature is its ability to send real messages and images through WeChat and QQ, meaning the agent doesn't just answer questions but actively reaches out to you in your own conversations. It reads your chats, understands the context, and replies like a person with her own memory and personality.

🆔 @hackernewsgithubprojects

📦 lean-dojo/torchlean

TorchLean: Proving Neural Networks

TorchLean is the neural network framework that lets you mathematically prove your models work. Instead of trusting code to do the right thing, you write the network and the proof in the same language called Lean. This means you can train a model and then formally verify it will never fail on certain inputs. It solves the problem of blind trust in artificial intelligence by turning predictions into checked facts. The result is a system where you can be absolutely certain your algorithm behaves as intended, which is huge for safety-critical applications.

🆔 @hackernewsgithubprojects

📦 femboyisp/yip

yip: The VPN That Fixes Packet Loss

Standard virtual private networks crumble the moment your Wi-Fi drops a single packet, cutting speeds by up to ninety-nine percent. yip fixes this by using a math trick called forward error correction to rebuild missing data before your computer even notices it is gone. Written in Rust, this tool turns your devices into a private, encrypted mesh that ignores unstable cellular and satellite connections. It does not hide your identity from peers, but it keeps traffic flowing even when the internet is broken. If you rely on unstable links, this is the fix you need.

🆔 @hackernewsgithubprojects

📦 liyupi/dafeiyu-cleaner

Dafeiyu Cleaner: The Browser Plugin That Hunts AI Bots

A cartoon whale character physically whips and knocks out AI bots right inside your browser comments. This project, called Dafeiyu Cleaner, is a browser extension for Chrome, Edge, and Firefox that hunts down anonymous bot accounts flooding Bilibili video sections with automated text. It works by reading the raw data hidden behind each comment, checking for robotic patterns, and even asking an AI model to double-check suspicious names. When it finds a fake account, the whale mascot appears, strikes the bot with a weapon, and hides the comment locally so you never see it again. It does not delete anything on the server; it just cleans your personal view.

🆔 @hackernewsgithubprojects

📦 chinaopenclaw/mcd-nutrition-expert

McDonald's Nutrition Expert: Eat Smart, Order Fast

McDonald's Nutrition Expert maps every menu item onto a strategic matrix to reveal exactly which foods fit your daily calorie and protein targets. It solves the confusing problem of choosing healthy fast food by calculating your personal metabolic needs and recommending specific meal combinations that stay within your sodium and energy limits. The system even connects to the official order service so you can buy your recommended meal in one click. This tool turns a chaotic menu into a clear, personalized plan for anyone managing their weight, proving you can enjoy fast food without guessing your nutrition.

🆔 @hackernewsgithubprojects

📦 tom123223/maimen-xinshi

Maimen Xinshi: The Emotional McDonald's Meal Planner

Maimen Xinshi is the emotional meal planner that turns your daily feelings into a real McDonald's order. It reads your mood to build a specific meal, then lets three distinct character types vote on whether you should actually eat it. This tool connects directly to live store data, so the prices and nutritional facts are accurate to the minute, not guessed. It helps you decide if a snack fits your current energy and emotional needs. It is a playful way to think about comfort food without feeling guilty. Give it a try for a unique dining perspective.

🆔 @hackernewsgithubprojects

📦 lillyguisnet/sam3-segmentation

Counting Corn Kernels With AI

One coding agent used a new tool to find 1,397 individual kernels on a single ear of corn in three minutes, and that is the specific thing this project does. Sam3-segmentation is a plug-in for coding assistants like Claude Code or Cursor that gives them access to Meta's image segmentation model. Instead of just guessing what is in a photo, the agent can actually trace the outlines of objects, count them, and save the results. The coolest part is that it treats the AI's guesses as drafts, not facts. It builds a review sheet so you can check the work before trusting the numbers.

🆔 @hackernewsgithubprojects

📦 agenticfabriq/mnemiq

Mnemiq: Text-to-SQL That Refuses to Guess

Mnemiq is an open-source engine that translates plain English into database queries, but it refuses to guess when your data lacks the answer. Unlike typical tools that might fabricate a plausible number, this system checks if the requested information actually exists in your tables. If a column for ticket revenue is missing, it tells you exactly what is missing instead of lying. You can tune every step, from the language model to the security rules, and measure the results on your specific database. This makes it a transparent tool for developers who need accurate data and clear reasoning.

📰 https://news.ycombinator.com/item?id=49626513

🆔 @hackernewsgithubprojects

📦 exmergo/dex

Dex: Your Agent's Data Safety Net

Dex stops your coding agent from accidentally leaking private data or spending your money by enforcing strict rules directly inside the engine. Most AI tools treat your database as a blank slate, constantly forgetting what they learned and blindly running expensive queries. Dex fixes this by acting as a careful librarian that checks the cost before scanning and blocks personal data from leaving the building. It explores your warehouse, builds your models, and warns you when things change, all while keeping your source data safe. If you use AI to touch your data, this is the guardrail you need to sleep well at night.

📰 https://news.ycombinator.com/item?id=49004816

🆔 @hackernewsgithubprojects

📦 rvdende/cadrs

cadrs

Build a full 3D part from a single flat sketch using a CAD app called cadrs. It runs entirely on your machine, written in Rust, so it feels snappy and local. The coolest part? You can plug an AI assistant directly into the app to draw, edit, and export your designs, and every change shows up in the history like you typed it yourself. It’s a serious tool for making real, printable parts without the cloud clutter. If you want to design hardware that actually works, this is the one to watch.

📰 https://news.ycombinator.com/item?id=50129079

🆔 @hackernewsgithubprojects

📦 dotnet/maui-labs

What is Maui Labs?

Let your AI assistant actually click buttons and fix bugs inside your mobile app. The Maui Labs collection includes a powerful automation tool that lets an AI agent see your interface, interact with elements, and diagnose problems in real time without you manually testing everything. It connects to the app's visual tree to stream logs and performance data directly to your coding assistant. This turns the messy process of mobile debugging into a collaborative task where the computer helps you verify fixes instantly. It is currently pre-release, so expect some changes, but it is a fascinating look at where developer tools are heading.

🆔 @hackernewsgithubprojects

📦 gaojimmy27/mcd-group-order

McD Group Order

McDonald’s now has an AI that splits the bill for you. This tool, called McD Group Order, fixes the awkwardness of figuring out who owes what after a group lunch. You just list what everyone ordered, and it crunches the numbers to give each person an exact amount, handling coupons and delivery fees so nobody overpays. It even connects directly to the restaurant system to place the order for the whole group. It is a clever way to automate the social friction of shared meals. Next time you go out, let the code handle the math.

🆔 @hackernewsgithubprojects

📦 redth/maui.sherpa

Maui Sherpa: Your Dev Environment Concierge

Tame your messy development environment with Maui Sherpa. This open source desktop app acts as a single control center for building and testing mobile applications, handling everything from managing your code tools to pushing apps to physical phones. Instead of hunting for obscure settings, you get one place to see your devices, manage digital keys, and check for errors. The coolest part is how it handles the boring stuff, like keeping your software updates organized, so you can focus on writing code. It’s a lifesaver for anyone tired of wrestling with setup. Grab it, let it fix your machine, and get back to building.

🆔 @hackernewsgithubprojects

📦 gputw-ai/gputw-skill

GPUtw Skill: Teach AI to Rent GPUs

GPUtw Skill is the AI knowledge package that turns your coding assistant into a competent cloud engineer. The problem it solves is simple: renting a GPU from a cloud provider usually requires memorizing dozens of API endpoints and complex authentication rules. This project fixes that by giving tools like Claude Code and Cursor a complete manual. Instead of guessing, your AI assistant now reads a structured guide to deploy machines, monitor their health, and move large model files into secure storage. It even includes a specialized server that lets the AI interact with the cloud using safe, typed commands rather than raw text.

🆔 @hackernewsgithubprojects

📦 miao-org/awesome-supermimic

SuperMimic: How Robots Learn to Dance

Gather every implementation of a robot learning system into one spot. That is what awesome SuperMimic does. It is a curated list of code, demos, and tools built by the community to help humanoid robots mimic human movement. The original method lets a robot figure out how to move its whole body by practicing in a simulator, then it tries those moves on real hardware. This repository collects all those efforts, from simple tests to complex setups. You can see how different groups tackled the same challenge, from dancing on flat floors to handling tricky terrain.

🆔 @hackernewsgithubprojects

📦 anonnerdfrenzy/survival-llm

Survival LLM: Your Offline AI Assistant

Survival LLM is the offline AI assistant that proves you don't need the internet to stay safe. It packs about 1.5 million real-world articles, from military first-aid manuals to Wikipedia's most-read pages, directly onto your computer. When the power grid fails and cell towers go dark, you can still ask it exactly how to purify water or treat a burn. The truly surprising part is that it refuses to guess; every single answer comes with a clickable link to the exact passage it read so you can verify the advice yourself.

🆔 @hackernewsgithubprojects