This media is not supported in your browser
VIEW IN TELEGRAM
π¦ itspatmorgan/design-studio
Design Studio: The Code-First Prototype Sandbox
Sketch a living interface where every button works and every page updates itself in real time. Design Studio is an open source workspace built for designers and product managers who actually code, giving them a place to build interactive prototypes without waiting for a development team. The standout feature is its module system, which lets you drop in custom components and see them instantly in a live preview, turning static mockups into functional demos. It solves the gap between design files and real software by letting you test flows, edit content, and refine layouts all in one place.
π @hackernewsgithubprojects
Design Studio: The Code-First Prototype Sandbox
Sketch a living interface where every button works and every page updates itself in real time. Design Studio is an open source workspace built for designers and product managers who actually code, giving them a place to build interactive prototypes without waiting for a development team. The standout feature is its module system, which lets you drop in custom components and see them instantly in a live preview, turning static mockups into functional demos. It solves the gap between design files and real software by letting you test flows, edit content, and refine layouts all in one place.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ mirage-hq/tesseract
Tesseract: Let Your AI Agent Direct Video
Your AI agent can now direct entire video scenes, complete with three dimensional camera moves and sound design. Tesseract by Mirage turns your coding assistant into a full creative editor. You simply describe the video you want, and the agent handles the cutting, motion graphics, and rendering. It runs locally on your machine, meaning your footage stays private while your agent iterates on the project. This works with popular tools like Chat G P T and Cursor. No complex interface needed, just talk to your agent. It is a fascinating shift from manual editing to conversational production.
π @hackernewsgithubprojects
Tesseract: Let Your AI Agent Direct Video
Your AI agent can now direct entire video scenes, complete with three dimensional camera moves and sound design. Tesseract by Mirage turns your coding assistant into a full creative editor. You simply describe the video you want, and the agent handles the cutting, motion graphics, and rendering. It runs locally on your machine, meaning your footage stays private while your agent iterates on the project. This works with popular tools like Chat G P T and Cursor. No complex interface needed, just talk to your agent. It is a fascinating shift from manual editing to conversational production.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ ywedai/teachable-code
Turn Any Codebase Into an Interactive 3D Lesson
Turn any codebase into a living, interactive lesson you can actually explore. Teachable Code is an agent skill that scans a repository and builds a single, self-contained page where the entire system stays visible on one screen. As you follow the data step by step, you see exactly how values transform, where they live in the source files, and how different parts connect, using three-dimensional views for complex data flows and two-dimensional layouts for applications. It turns abstract code into a tangible map you can inspect, helping you understand the hidden logic and real behavior of a project in plain terms.
π @hackernewsgithubprojects
Turn Any Codebase Into an Interactive 3D Lesson
Turn any codebase into a living, interactive lesson you can actually explore. Teachable Code is an agent skill that scans a repository and builds a single, self-contained page where the entire system stays visible on one screen. As you follow the data step by step, you see exactly how values transform, where they live in the source files, and how different parts connect, using three-dimensional views for complex data flows and two-dimensional layouts for applications. It turns abstract code into a tangible map you can inspect, helping you understand the hidden logic and real behavior of a project in plain terms.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ darhkfox/supermarioallstars
Super Mario All-Stars: True Widescreen & Co-Op
A homebrew port of Super Mario All-Stars lets you play the classic SNES collection on Windows and Android with true widescreen support up to twenty-one to nine aspect ratios. The most surprising feature is simultaneous two-player co-op for the first three games, allowing Mario and Luigi to play together in real-time with shared lives and independent physics. This native application also includes a launcher with live previews, eighty-two achievements, a music player, and CRT filters that make the retro visuals look beautiful on modern displays. It requires your own USA ROM file to run, ensuring you own the original game.
π @hackernewsgithubprojects
Super Mario All-Stars: True Widescreen & Co-Op
A homebrew port of Super Mario All-Stars lets you play the classic SNES collection on Windows and Android with true widescreen support up to twenty-one to nine aspect ratios. The most surprising feature is simultaneous two-player co-op for the first three games, allowing Mario and Luigi to play together in real-time with shared lives and independent physics. This native application also includes a launcher with live previews, eighty-two achievements, a music player, and CRT filters that make the retro visuals look beautiful on modern displays. It requires your own USA ROM file to run, ensuring you own the original game.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ lexifina/romania-law
Romania Law
Every single law in Romania is stored inside a standard git repository, meaning the entire legal history of a nation lives in version control. Instead of scrolling through endless official portals, you can track exactly how any article changed over decades using simple command line tools. This project rebuilds the Constitution, codes, and court rulings into one searchable timeline where every amendment is a distinct commit. It turns complex legislation into clear, traceable data that anyone can inspect. For developers and researchers, it makes Romanian law transparent and easy to analyze. The takeaway? Git isn't just for code, it can hold the rules of a country.
π @hackernewsgithubprojects
Romania Law
Every single law in Romania is stored inside a standard git repository, meaning the entire legal history of a nation lives in version control. Instead of scrolling through endless official portals, you can track exactly how any article changed over decades using simple command line tools. This project rebuilds the Constitution, codes, and court rulings into one searchable timeline where every amendment is a distinct commit. It turns complex legislation into clear, traceable data that anyone can inspect. For developers and researchers, it makes Romanian law transparent and easy to analyze. The takeaway? Git isn't just for code, it can hold the rules of a country.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ deepseek-ai/engram
Engram: Smarter Memory for AI
Engram is the memory module that finally gives large language models a direct line to their own knowledge. DeepSeek built it to solve a stubborn problem: standard AI models have to guess at stored facts, which is slow and error-prone. Engram acts like a lightning-fast lookup table, letting the model instantly retrieve specific information without doing heavy math every time. This frees up brainpower for harder reasoning tasks, making the system faster and more accurate. It is a clever way to make AI smarter without just adding more processing power. If you love watching machines learn efficiently, this is a cool piece of the puzzle to keep in your pocket.
π @hackernewsgithubprojects
Engram: Smarter Memory for AI
Engram is the memory module that finally gives large language models a direct line to their own knowledge. DeepSeek built it to solve a stubborn problem: standard AI models have to guess at stored facts, which is slow and error-prone. Engram acts like a lightning-fast lookup table, letting the model instantly retrieve specific information without doing heavy math every time. This frees up brainpower for harder reasoning tasks, making the system faster and more accurate. It is a clever way to make AI smarter without just adding more processing power. If you love watching machines learn efficiently, this is a cool piece of the puzzle to keep in your pocket.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ shanselman/sandbox-mcp
Test Your Windows Apps on a Disposable PC
Sandbox MCP boots a fresh, throwaway Windows machine in seconds so you can verify your app actually works for a real user. It solves the classic problem where software runs perfectly on your dev box but crashes on a customerβs computer because they are missing a specific runtime or system file. The tool boots a clean, offline desktop environment, deploys your application, drives the user interface to catch crashes, and captures screenshots as proof. Once the test is done, the entire virtual machine is instantly deleted, leaving you with a clean report and zero leftover mess on your actual computer.
π @hackernewsgithubprojects
Test Your Windows Apps on a Disposable PC
Sandbox MCP boots a fresh, throwaway Windows machine in seconds so you can verify your app actually works for a real user. It solves the classic problem where software runs perfectly on your dev box but crashes on a customerβs computer because they are missing a specific runtime or system file. The tool boots a clean, offline desktop environment, deploys your application, drives the user interface to catch crashes, and captures screenshots as proof. Once the test is done, the entire virtual machine is instantly deleted, leaving you with a clean report and zero leftover mess on your actual computer.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ jaigner-hub/ocarina-of-time-recomp
Zelda: Ocarina of Time PC Recomp
The Legend of Zelda Ocarina of Time PC Recomp is the native Windows build that finally runs this Switch 2 title directly on your computer without a console. The repository takes the game's original code, lifts it into standard C, and links it with a custom runtime that replaces the hardware with standard PC graphics, making the game run natively on modern processors. The most surprising part is how it handles the title screen cinematic.
π @hackernewsgithubprojects
Zelda: Ocarina of Time PC Recomp
The Legend of Zelda Ocarina of Time PC Recomp is the native Windows build that finally runs this Switch 2 title directly on your computer without a console. The repository takes the game's original code, lifts it into standard C, and links it with a custom runtime that replaces the hardware with standard PC graphics, making the game run natively on modern processors. The most surprising part is how it handles the title screen cinematic.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ deeix-ai/winc
Winc: The Editor That Starts in 0.1 Seconds
Open a hundred thousand lines of code without a blink, thanks to Winc. This isn't your average text editor; it is a desktop powerhouse built entirely from scratch in Rust. The real magic is its own rendering engine, which means it boots in just one-tenth of a second and sips only fifty-two megabytes of memory while handling massive files. It feels less like software and more like a direct line to your machine. If you are tired of waiting for your tools to catch up, this is the one to try. It is fast, lean, and actually enjoyable to use.
π @hackernewsgithubprojects
Winc: The Editor That Starts in 0.1 Seconds
Open a hundred thousand lines of code without a blink, thanks to Winc. This isn't your average text editor; it is a desktop powerhouse built entirely from scratch in Rust. The real magic is its own rendering engine, which means it boots in just one-tenth of a second and sips only fifty-two megabytes of memory while handling massive files. It feels less like software and more like a direct line to your machine. If you are tired of waiting for your tools to catch up, this is the one to try. It is fast, lean, and actually enjoyable to use.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ lijiaqi-ai-product-manager/dsh-agent-compare
AgentCompare: Build Two Apps, Pick One
AgentCompare gives you a split screen where two different coding agents work on the exact same frontend task at the same time. Instead of guessing which version of your code is better, the tool spins up two separate, live versions of your application so you can actually click around and feel the difference. It runs real checks on both builds to make sure they work before you decide. You only find out which strategy or model created each version after you have made your choice, keeping the test fair. Once you pick your favorite, it merges that specific code back into your project.
π @hackernewsgithubprojects
AgentCompare: Build Two Apps, Pick One
AgentCompare gives you a split screen where two different coding agents work on the exact same frontend task at the same time. Instead of guessing which version of your code is better, the tool spins up two separate, live versions of your application so you can actually click around and feel the difference. It runs real checks on both builds to make sure they work before you decide. You only find out which strategy or model created each version after you have made your choice, keeping the test fair. Once you pick your favorite, it merges that specific code back into your project.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ oz9un/sshole
Sshole: The 257 Character Password Bug
Log into root on an Alpine server with a random two hundred and fifty seven character password. Sshole reveals a decade-old flaw where OpenSSH running on Musl Linux accepted any long enough password for accounts locked with an asterisk. It happens because the system confused a failure signal with a real hash. If you run Alpine, OpenWRT, or Buildroot, you might have been wide open since two thousand sixteen. This repo gives you a simple script to check your machine, a Docker setup to test the exploit safely, and the official patch.
π @hackernewsgithubprojects
Sshole: The 257 Character Password Bug
Log into root on an Alpine server with a random two hundred and fifty seven character password. Sshole reveals a decade-old flaw where OpenSSH running on Musl Linux accepted any long enough password for accounts locked with an asterisk. It happens because the system confused a failure signal with a real hash. If you run Alpine, OpenWRT, or Buildroot, you might have been wide open since two thousand sixteen. This repo gives you a simple script to check your machine, a Docker setup to test the exploit safely, and the official patch.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ glen-web-app/best-agent-memory-mcp
Best Agent Memory MCP
The Best Agent Memory MCP is the comparison list that finally cuts through the noise of AI tools that forget everything. It ranks twenty-five memory servers that let your coding assistants remember facts and decisions across sessions so you never have to repeat yourself. The real surprise here is how it separates hosted services like Glen and Mem0 from open-source options you run yourself, giving you clear answers on storage and team sharing. No more guessing which tool actually works with your setup. If you are building with AI agents, this single resource tells you exactly which memory layer fits your needs and which ones to skip entirely.
π @hackernewsgithubprojects
Best Agent Memory MCP
The Best Agent Memory MCP is the comparison list that finally cuts through the noise of AI tools that forget everything. It ranks twenty-five memory servers that let your coding assistants remember facts and decisions across sessions so you never have to repeat yourself. The real surprise here is how it separates hosted services like Glen and Mem0 from open-source options you run yourself, giving you clear answers on storage and team sharing. No more guessing which tool actually works with your setup. If you are building with AI agents, this single resource tells you exactly which memory layer fits your needs and which ones to skip entirely.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ vlcekm/meitouclient
Run Kenshi on Mac
Run Kenshi on your Mac. That is the whole point of Meitou Client. This open source project rebuilds the game's engine from scratch in modern C sharp. It stops relying on the original, locked down Windows code and instead uses a custom Vulkan renderer. This means the game actually works on macOS and other systems where the official client fails. You still need your own copy of the game files, but the engine doing the heavy lifting is completely new. It is a massive amount of work to translate every piece of graphics and logic into this new framework.
π @hackernewsgithubprojects
Run Kenshi on Mac
Run Kenshi on your Mac. That is the whole point of Meitou Client. This open source project rebuilds the game's engine from scratch in modern C sharp. It stops relying on the original, locked down Windows code and instead uses a custom Vulkan renderer. This means the game actually works on macOS and other systems where the official client fails. You still need your own copy of the game files, but the engine doing the heavy lifting is completely new. It is a massive amount of work to translate every piece of graphics and logic into this new framework.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ nico-bohlinger/xmimic
Train a Dancing Robot in 60 Seconds
Teach a human-sized robot to dance in under sixty seconds. XMimic is a new tool that trains a Unitree G1 robot to copy human movements, and it does so three and a half times faster than existing methods. By splitting the heavy math across four powerful graphics cards, it finishes learning in just ninety-one seconds. This speed means researchers can test new dance styles without waiting hours for the computer to figure them out. The result is the same smooth, fluid motion, but the training process is nearly instant. If you work with robot learning, this is the shortcut you have been waiting for to speed up your experiments.
π @hackernewsgithubprojects
Train a Dancing Robot in 60 Seconds
Teach a human-sized robot to dance in under sixty seconds. XMimic is a new tool that trains a Unitree G1 robot to copy human movements, and it does so three and a half times faster than existing methods. By splitting the heavy math across four powerful graphics cards, it finishes learning in just ninety-one seconds. This speed means researchers can test new dance styles without waiting hours for the computer to figure them out. The result is the same smooth, fluid motion, but the training process is nearly instant. If you work with robot learning, this is the shortcut you have been waiting for to speed up your experiments.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ kinference-ai/kinference
Kinference: Smarter LLM Routing
Kinference solves a subtle problem that most people miss when running large language models at scale. It knows exactly which parts of your prompt are already cached on which server, meaning it can route your request to the machine that will answer fastest. Instead of guessing based on simple load, it looks at the actual data sitting in memory. This makes your API calls significantly faster and cheaper by avoiding redundant work. It is a clever way to manage the messy reality of distributed AI workloads. If you are running models on Kubernetes, this is the kind of detail that separates a slow demo from a fast product.
π @hackernewsgithubprojects
Kinference: Smarter LLM Routing
Kinference solves a subtle problem that most people miss when running large language models at scale. It knows exactly which parts of your prompt are already cached on which server, meaning it can route your request to the machine that will answer fastest. Instead of guessing based on simple load, it looks at the actual data sitting in memory. This makes your API calls significantly faster and cheaper by avoiding redundant work. It is a clever way to manage the messy reality of distributed AI workloads. If you are running models on Kubernetes, this is the kind of detail that separates a slow demo from a fast product.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ tiger-ai-lab/websitebench
WebsiteBench: The AI Website Test
An AI agent can now look at a live website and build a working copy of it just by browsing, without reading a single line of code. WebsiteBench is a new benchmark from the Tiger A I Lab that tests whether large language models can truly understand and reconstruct web interfaces through visual exploration. The project provides a suite of real-world tasks, ranging from simple article pages to complex e-commerce sites like Amazon and Airtable, where agents must replicate the site's structure and content. This matters because it reveals if A I can actually comprehend the layout and logic of the web, not just generate text.
π° https://news.ycombinator.com/item?id=50378231
π @hackernewsgithubprojects
WebsiteBench: The AI Website Test
An AI agent can now look at a live website and build a working copy of it just by browsing, without reading a single line of code. WebsiteBench is a new benchmark from the Tiger A I Lab that tests whether large language models can truly understand and reconstruct web interfaces through visual exploration. The project provides a suite of real-world tasks, ranging from simple article pages to complex e-commerce sites like Amazon and Airtable, where agents must replicate the site's structure and content. This matters because it reveals if A I can actually comprehend the layout and logic of the web, not just generate text.
π° https://news.ycombinator.com/item?id=50378231
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ streamingtelligence/gelato
Gelato: The Protocol Giving AI Hands
Gelato is an open standard that lets an artificial intelligence take direct control of your physical phone, camera, or smart glasses without a single cable. It works like a secure bridge where the AI can see what your device sees and send commands to tap, type, or stream video, all over a normal Wi-Fi network. This matters because it turns your existing hardware into a live assistant that can actually help you in the real world, not just chat. The system keeps your private data safe by never sending your login passwords to the cloud.
π @hackernewsgithubprojects
Gelato: The Protocol Giving AI Hands
Gelato is an open standard that lets an artificial intelligence take direct control of your physical phone, camera, or smart glasses without a single cable. It works like a secure bridge where the AI can see what your device sees and send commands to tap, type, or stream video, all over a normal Wi-Fi network. This matters because it turns your existing hardware into a live assistant that can actually help you in the real world, not just chat. The system keeps your private data safe by never sending your login passwords to the cloud.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ oil-oil/oil-diff
Oil-diff: Your Code Changes, Beautifully Illustrated
Oil-diff is the tool that turns messy code updates into a beautiful, readable report. Instead of staring at dry text logs or confusing terminal output, it builds a polished webpage that highlights exactly what changed in your work. This makes it much easier to see the big picture and share progress with teammates. It is a handy way to make your technical work more accessible and visually engaging. Give Oil-diff a try to see your code changes in a whole new light.
π @hackernewsgithubprojects
Oil-diff: Your Code Changes, Beautifully Illustrated
Oil-diff is the tool that turns messy code updates into a beautiful, readable report. Instead of staring at dry text logs or confusing terminal output, it builds a polished webpage that highlights exactly what changed in your work. This makes it much easier to see the big picture and share progress with teammates. It is a handy way to make your technical work more accessible and visually engaging. Give Oil-diff a try to see your code changes in a whole new light.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ camthink-ai/neomind
NeoMind: Your Local AI Home Hub
NeoMind lets you talk to your IoT devices like they have a brain. Built in Rust for speed, it runs local AI models directly on your machine, so your data never leaves your house. The coolest part? You can just ask it to do things. Want your lights on when the sensor trips? NeoMind sets that up automatically. Itβs a full smart home hub that actually thinks, not just reacts. No cloud fees, no privacy leaks, just pure local intelligence. If you want to take control of your smart home, this is the project to watch.
π @hackernewsgithubprojects
NeoMind: Your Local AI Home Hub
NeoMind lets you talk to your IoT devices like they have a brain. Built in Rust for speed, it runs local AI models directly on your machine, so your data never leaves your house. The coolest part? You can just ask it to do things. Want your lights on when the sensor trips? NeoMind sets that up automatically. Itβs a full smart home hub that actually thinks, not just reacts. No cloud fees, no privacy leaks, just pure local intelligence. If you want to take control of your smart home, this is the project to watch.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ dis0rder0x00/trustedhive
TrustedHive: How a Normal User Became SYSTEM on Windows 11
A normal Windows user account, with no admin rights and no special permissions, can now force the operating system to write any file anywhere you want, owned by the most privileged account in the system. This is a local privilege escalation bug in Windows 11 that hides inside the update servicing stack. The trick involves tricking a system process called TrustedInstaller into following a fake file path that you create, allowing your low-level code to slip a malicious file into protected system folders. It was patched in September, but the proof of concept code is public and shows exactly how the hole worked.
π @hackernewsgithubprojects
TrustedHive: How a Normal User Became SYSTEM on Windows 11
A normal Windows user account, with no admin rights and no special permissions, can now force the operating system to write any file anywhere you want, owned by the most privileged account in the system. This is a local privilege escalation bug in Windows 11 that hides inside the update servicing stack. The trick involves tricking a system process called TrustedInstaller into following a fake file path that you create, allowing your low-level code to slip a malicious file into protected system folders. It was patched in September, but the proof of concept code is public and shows exactly how the hole worked.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ roycho87/geiru_context_suite
Geiru Context Suite Keeps Long AI Videos From Falling Apart
Geiru Context Suite stops the visual quality of long AI generated videos from slowly falling apart. When you stitch together multiple short video clips to make a longer scene the model loses detail with every new piece it generates. This tool solves that problem by locking the first clean frame of your video into every subsequent clip. It acts like an anchor pulling the image back to its original sharpness so colors and textures do not degrade over time. You can even turn this protection on or off with a single switch depending on your scene changes.
π @hackernewsgithubprojects
Geiru Context Suite Keeps Long AI Videos From Falling Apart
Geiru Context Suite stops the visual quality of long AI generated videos from slowly falling apart. When you stitch together multiple short video clips to make a longer scene the model loses detail with every new piece it generates. This tool solves that problem by locking the first clean frame of your video into every subsequent clip. It acts like an anchor pulling the image back to its original sharpness so colors and textures do not degrade over time. You can even turn this protection on or off with a single switch depending on your scene changes.
π @hackernewsgithubprojects