This media is not supported in your browser
VIEW IN TELEGRAM
π¦ facebookresearch/swe-sweep
SWE-sweep: The Bug Hunter Benchmark
SWE-sweep gives AI agents a real codebase and lets them hunt down bugs on their own, with zero hints about where to look. It is a stress test for large language models, asking them to find and fix as many issues as possible in complex software. The interesting part is that the AI has to figure out what is broken without any guidance, just like a human developer debugging a tricky app. This project, built by Facebook Research, helps us understand how well AI can actually handle real-world coding problems instead of just simple exercises.
π @hackernewsgithubprojects
SWE-sweep: The Bug Hunter Benchmark
SWE-sweep gives AI agents a real codebase and lets them hunt down bugs on their own, with zero hints about where to look. It is a stress test for large language models, asking them to find and fix as many issues as possible in complex software. The interesting part is that the AI has to figure out what is broken without any guidance, just like a human developer debugging a tricky app. This project, built by Facebook Research, helps us understand how well AI can actually handle real-world coding problems instead of just simple exercises.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ gillespiejameson/personal-finance-tracker
Personal Finance Tracker
Drop a raw bank statement into Personal Finance Tracker and watch it automatically identify the bank, fix the dates, and merge the data into one clean searchable table. This local first tool keeps your entire financial history in a single file on your own computer, so there are no cloud servers or accounts to manage. It goes far beyond a simple ledger by spotting recurring bills and paychecks, then calculating a real safe to spend number alongside your total net worth. You also get cash flow forecasts and anomaly alerts that warn you when spending suddenly shifts out of your normal pattern.
π @hackernewsgithubprojects
Personal Finance Tracker
Drop a raw bank statement into Personal Finance Tracker and watch it automatically identify the bank, fix the dates, and merge the data into one clean searchable table. This local first tool keeps your entire financial history in a single file on your own computer, so there are no cloud servers or accounts to manage. It goes far beyond a simple ledger by spotting recurring bills and paychecks, then calculating a real safe to spend number alongside your total net worth. You also get cash flow forecasts and anomaly alerts that warn you when spending suddenly shifts out of your normal pattern.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ boxabirds/awesome-local-ai
awesome-local-ai: The One-Command Way to Run Real AI Locally
Get a powerful AI assistant running on your own hardware in just one command. This project cuts through the endless setup tutorials by giving you ready-made installers that automatically pick the best model for your specific machine. The coolest part is that they actually test these setups to see if the AI can build real software, so you know it works. No guessing, no broken configs. Just one script that finds your perfect fit and gets you coding with local intelligence immediately. Itβs the shortcut to private, offline AI that actually respects your hardware. Skip the setup pain and start creating with a tool thatβs ready to go.
π @hackernewsgithubprojects
awesome-local-ai: The One-Command Way to Run Real AI Locally
Get a powerful AI assistant running on your own hardware in just one command. This project cuts through the endless setup tutorials by giving you ready-made installers that automatically pick the best model for your specific machine. The coolest part is that they actually test these setups to see if the AI can build real software, so you know it works. No guessing, no broken configs. Just one script that finds your perfect fit and gets you coding with local intelligence immediately. Itβs the shortcut to private, offline AI that actually respects your hardware. Skip the setup pain and start creating with a tool thatβs ready to go.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ dddevid/spotlight
SpotLight: Spotify without the bloat
The official Spotify desktop app weighs in at over four hundred megabytes and often eats gigabytes of memory, but SpotLight proves you don't need a web browser to listen to your music. Built entirely in Rust, this ultra-lightweight client replaces the heavy Electron wrapper with native code that runs on just sixty megabytes of RAM. It connects directly to your Premium account to play audio, search tracks, and scroll through synced lyrics without the background bloat that drains your laptop battery. When you aren't interacting with the window, the interface completely goes to sleep, keeping CPU usage near zero.
π @hackernewsgithubprojects
SpotLight: Spotify without the bloat
The official Spotify desktop app weighs in at over four hundred megabytes and often eats gigabytes of memory, but SpotLight proves you don't need a web browser to listen to your music. Built entirely in Rust, this ultra-lightweight client replaces the heavy Electron wrapper with native code that runs on just sixty megabytes of RAM. It connects directly to your Premium account to play audio, search tracks, and scroll through synced lyrics without the background bloat that drains your laptop battery. When you aren't interacting with the window, the interface completely goes to sleep, keeping CPU usage near zero.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ morrowmake/glm53-flash-cmp170hx-recipe
GLM 5.3 Flash Recipe
A 320 billion parameter model is now served locally on just four mid-range graphics cards, pushing nearly a thousand tokens per second for eight simultaneous users. This repository, known as the GLM 5.3 Flash recipe, solves the problem of running massive AI models on hardware that was never officially designed for them. It works by using a specialized engine that adapts to specific graphics chip architecture, allowing a single command to install and launch the entire system in an isolated container. The real surprise is that it achieves this speed without cutting corners on data precision or offloading memory to your disk.
π @hackernewsgithubprojects
GLM 5.3 Flash Recipe
A 320 billion parameter model is now served locally on just four mid-range graphics cards, pushing nearly a thousand tokens per second for eight simultaneous users. This repository, known as the GLM 5.3 Flash recipe, solves the problem of running massive AI models on hardware that was never officially designed for them. It works by using a specialized engine that adapts to specific graphics chip architecture, allowing a single command to install and launch the entire system in an isolated container. The real surprise is that it achieves this speed without cutting corners on data precision or offloading memory to your disk.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ eliovp-bv/paiton-vllm-plugin
Paiton: Run AI on AMD Radeon
Paiton is the AMD GPU plugin that finally lets you run heavy AI models locally without a proprietary cloud. It works by compiling your specific AI model directly into optimized code for your AMD Radeon graphics card, which means faster speeds and better privacy for your data. You can run everything from text chats and coding assistants to image and video generators, all on your own hardware. Whether you are a developer testing a new model or a creator looking for local tools, this gives you total control over your AI stack.
π @hackernewsgithubprojects
Paiton: Run AI on AMD Radeon
Paiton is the AMD GPU plugin that finally lets you run heavy AI models locally without a proprietary cloud. It works by compiling your specific AI model directly into optimized code for your AMD Radeon graphics card, which means faster speeds and better privacy for your data. You can run everything from text chats and coding assistants to image and video generators, all on your own hardware. Whether you are a developer testing a new model or a creator looking for local tools, this gives you total control over your AI stack.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ pangular-inspector/devtools
Devtools
Poke inside a live Angular app to see exactly how its pieces fit together. That is what Devtools does. It runs as a Chrome extension or a command line tool to inspect your component tree, signals, and routes right inside your browser. It is not just for humans, though. It can also feed this live data directly to your AI coding agents. This means your AI assistant can actually see your running app while it helps you write code. It turns invisible internal state into plain, readable data. No more guessing what your app is doing. Just clear, live answers.
π @hackernewsgithubprojects
Devtools
Poke inside a live Angular app to see exactly how its pieces fit together. That is what Devtools does. It runs as a Chrome extension or a command line tool to inspect your component tree, signals, and routes right inside your browser. It is not just for humans, though. It can also feed this live data directly to your AI coding agents. This means your AI assistant can actually see your running app while it helps you write code. It turns invisible internal state into plain, readable data. No more guessing what your app is doing. Just clear, live answers.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ avbiswas/bev-train
BEV Train
Build a decision model that picks the right answer no matter how you shuffle the options. Bev Train teaches a tiny neural network to score multiple choices using a large language model as its foundation. The clever trick? Each option gets its own secret lane, so the model evaluates every answer in isolation. This means the score stays the same whether the correct choice is first or last. Itβs a clean, efficient way to teach machines stable logic without the usual confusion. Perfect for anyone curious about how AI handles decision-making under pressure. Now you can train your own order-proof model in just a few steps.
π @hackernewsgithubprojects
BEV Train
Build a decision model that picks the right answer no matter how you shuffle the options. Bev Train teaches a tiny neural network to score multiple choices using a large language model as its foundation. The clever trick? Each option gets its own secret lane, so the model evaluates every answer in isolation. This means the score stays the same whether the correct choice is first or last. Itβs a clean, efficient way to teach machines stable logic without the usual confusion. Perfect for anyone curious about how AI handles decision-making under pressure. Now you can train your own order-proof model in just a few steps.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ slybase/vllm-sly-radiance
Run 27B AI on One Radeon Card
Serve a massive twenty seven billion parameter language model directly from a single thirty two gigabyte AMD Radeon Pro card. This project, called VLLM Sly Radiance, packages a highly tuned inference engine that squeezes maximum speed out of consumer level hardware without needing a server rack. The standout feature is a specialized kernel that processes math with four bit precision, allowing the model to read and write data significantly faster than standard methods. This pushes single stream decoding speeds to over one hundred thirty five tokens per second while maintaining a massive context window.
π @hackernewsgithubprojects
Run 27B AI on One Radeon Card
Serve a massive twenty seven billion parameter language model directly from a single thirty two gigabyte AMD Radeon Pro card. This project, called VLLM Sly Radiance, packages a highly tuned inference engine that squeezes maximum speed out of consumer level hardware without needing a server rack. The standout feature is a specialized kernel that processes math with four bit precision, allowing the model to read and write data significantly faster than standard methods. This pushes single stream decoding speeds to over one hundred thirty five tokens per second while maintaining a massive context window.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ maosika-ai/reshot
ReShot: Copy the Shot, Not the Actors
A single command line tool can extract the exact choreography and camera moves from a video clip, completely stripping out the original actors and faces. This is a project called ReShot. It solves a major problem in AI video generation where models usually copy the look of a reference clip along with its movement, or fail entirely when you try to describe the action in words. ReShot converts your source video into a clean control video, like a depth map or a skeleton, which feeds directly into video models.
π @hackernewsgithubprojects
ReShot: Copy the Shot, Not the Actors
A single command line tool can extract the exact choreography and camera moves from a video clip, completely stripping out the original actors and faces. This is a project called ReShot. It solves a major problem in AI video generation where models usually copy the look of a reference clip along with its movement, or fail entirely when you try to describe the action in words. ReShot converts your source video into a clean control video, like a depth map or a skeleton, which feeds directly into video models.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ rishilahoti/ashby-job-scraper
Ashby Job Scraper: 70,000 Jobs in One Feed
Pulling seventy thousand open roles from two thousand companies straight off their career pages is what Ashby Job Scraper does every single day. It solves the exhausting problem of checking ten different job boards for the same kind of work by scraping them all at once, automatically finding new companies, and keeping the list fresh without you lifting a finger. The most interesting part is how it ranks everything based on what you actually want, filtering out the noise so you only see roles that match your tech stack and location. It is a free, self-hosted tool that turns the chaos of the job market into one clean, sorted feed.
π @hackernewsgithubprojects
Ashby Job Scraper: 70,000 Jobs in One Feed
Pulling seventy thousand open roles from two thousand companies straight off their career pages is what Ashby Job Scraper does every single day. It solves the exhausting problem of checking ten different job boards for the same kind of work by scraping them all at once, automatically finding new companies, and keeping the list fresh without you lifting a finger. The most interesting part is how it ranks everything based on what you actually want, filtering out the noise so you only see roles that match your tech stack and location. It is a free, self-hosted tool that turns the chaos of the job market into one clean, sorted feed.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ druthulu/projectarchitect
Project Architect turns Claude Code into a team
Project Architect turns Claude Code into a small, coordinated team that actually follows a written plan. Instead of one giant chat window that forgets everything after an hour, it breaks your project into phases, assigns specialized AI agents to specific jobs like coding or research, and saves the progress to your local files. This means every new session starts fresh but remembers exactly where you left off, saving you massive amounts of money and time. It is basically a project manager for your AI, ensuring the work gets done without you holding the hand of the algorithm.
π @hackernewsgithubprojects
Project Architect turns Claude Code into a team
Project Architect turns Claude Code into a small, coordinated team that actually follows a written plan. Instead of one giant chat window that forgets everything after an hour, it breaks your project into phases, assigns specialized AI agents to specific jobs like coding or research, and saves the progress to your local files. This means every new session starts fresh but remembers exactly where you left off, saving you massive amounts of money and time. It is basically a project manager for your AI, ensuring the work gets done without you holding the hand of the algorithm.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ github/copilot-app-for-beginners
How to Direct AI Coding Agents in One Place
GitHub built a desktop app that turns your computer into a control center for directing multiple A.I. coding agents at once. Instead of asking a chatbot for a snippet, you give these agents a goal, watch them plan and write the code, and then verify the result before accepting it. The GitHub Copilot app for Beginners course teaches you how to use this single window to manage entire projects, test the output, and move finished work into pull requests. It solves the problem of jumping between five different tools by putting the agent, the code, and the review steps into one shared space.
π @hackernewsgithubprojects
How to Direct AI Coding Agents in One Place
GitHub built a desktop app that turns your computer into a control center for directing multiple A.I. coding agents at once. Instead of asking a chatbot for a snippet, you give these agents a goal, watch them plan and write the code, and then verify the result before accepting it. The GitHub Copilot app for Beginners course teaches you how to use this single window to manage entire projects, test the output, and move finished work into pull requests. It solves the problem of jumping between five different tools by putting the agent, the code, and the review steps into one shared space.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ opensensenova/looped-dit
Looped-DiT: The AI Image Model That Gets Smarter by Repeating Itself
Looped-DiT is the image generator that gets smarter by doing the same thing over and over. Instead of building a massive model with millions of extra parts, this system takes a small set of tools and runs them in a loop, like checking your work multiple times to get it right. It uses a clever trick to keep the process stable so the extra effort actually improves the picture. This means you can generate crisp images without needing a computer cluster. It is a neat way to boost quality through repetition rather than raw size. If you love seeing clever engineering, this one is worth a look.
π @hackernewsgithubprojects
Looped-DiT: The AI Image Model That Gets Smarter by Repeating Itself
Looped-DiT is the image generator that gets smarter by doing the same thing over and over. Instead of building a massive model with millions of extra parts, this system takes a small set of tools and runs them in a loop, like checking your work multiple times to get it right. It uses a clever trick to keep the process stable so the extra effort actually improves the picture. This means you can generate crisp images without needing a computer cluster. It is a neat way to boost quality through repetition rather than raw size. If you love seeing clever engineering, this one is worth a look.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ rexmhall09/tuff
TUFF: Run Giant AI on a Tiny Mac
TUFF runs a one-hundred-and-ten-gigabyte language model on a sixteen-gigabyte MacBook Air. Most local AI apps require the entire model to fit in your computerβs memory, but TUFF works around this by keeping most of the model on the hard drive. It streams only the specific parts the computer needs at that exact moment into memory. This is a native Mac application written in Swift and Metal, meaning it is not a wrapper around existing tools. It handles large models that would normally be impossible to run locally on smaller laptops. While it is slower than using a smaller model, it proves that powerful AI does not need a data center.
π @hackernewsgithubprojects
TUFF: Run Giant AI on a Tiny Mac
TUFF runs a one-hundred-and-ten-gigabyte language model on a sixteen-gigabyte MacBook Air. Most local AI apps require the entire model to fit in your computerβs memory, but TUFF works around this by keeping most of the model on the hard drive. It streams only the specific parts the computer needs at that exact moment into memory. This is a native Mac application written in Swift and Metal, meaning it is not a wrapper around existing tools. It handles large models that would normally be impossible to run locally on smaller laptops. While it is slower than using a smaller model, it proves that powerful AI does not need a data center.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ preppie22/gogstash
GogStash: The GOG Game Backup Tool
GogStash is the desktop app that finally makes backing up your entire GOG game library painless. It solves the headache of trying to save your digital games locally by offering a simple interface to grab every installer, patch, and piece of bonus content you own. The coolest part is how smart it is; it checks your hard drive first so it never wastes bandwidth downloading files you already have. You just log in, pick your games, and let it work in the background.
π @hackernewsgithubprojects
GogStash: The GOG Game Backup Tool
GogStash is the desktop app that finally makes backing up your entire GOG game library painless. It solves the headache of trying to save your digital games locally by offering a simple interface to grab every installer, patch, and piece of bonus content you own. The coolest part is how smart it is; it checks your hard drive first so it never wastes bandwidth downloading files you already have. You just log in, pick your games, and let it work in the background.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ aetherlabsai/causalwm
CausalWM: How AI Predicts Video
CausalWM predicts what a robot will do next by explicitly calculating the physics behind it. Instead of guessing pixels, the model first generates the motion and three-dimensional geometry of the scene before creating the actual video. This means you simply show it a picture and give it a command, and it figures out how objects move in space. It is interesting because it mimics how we understand cause and effect. By reasoning through these steps, the system creates videos that follow real-world logic. This makes it a powerful tool for training robots to interact safely with their environment. The result is a smarter way to simulate physical actions.
π @hackernewsgithubprojects
CausalWM: How AI Predicts Video
CausalWM predicts what a robot will do next by explicitly calculating the physics behind it. Instead of guessing pixels, the model first generates the motion and three-dimensional geometry of the scene before creating the actual video. This means you simply show it a picture and give it a command, and it figures out how objects move in space. It is interesting because it mimics how we understand cause and effect. By reasoning through these steps, the system creates videos that follow real-world logic. This makes it a powerful tool for training robots to interact safely with their environment. The result is a smarter way to simulate physical actions.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ devagrawal09/solid-yield
Solid Yield Makes Your TypeScript Lie
Turn your Solid code into a type system that actually knows what can break. Solid Yield rewrites your components into generator functions so that every possible failure, pending state, or missing context becomes a visible part of your function signature. This solves the annoying problem where your types claim a component is safe, but it crashes the moment a network request fails. It is wild because the compiler literally refuses to let you forget a loading spinner or error handler. If your types pass, your app won't break. It is a strict, clever way to guarantee your UI handles every edge case before you even run it.
π @hackernewsgithubprojects
Solid Yield Makes Your TypeScript Lie
Turn your Solid code into a type system that actually knows what can break. Solid Yield rewrites your components into generator functions so that every possible failure, pending state, or missing context becomes a visible part of your function signature. This solves the annoying problem where your types claim a component is safe, but it crashes the moment a network request fails. It is wild because the compiler literally refuses to let you forget a loading spinner or error handler. If your types pass, your app won't break. It is a strict, clever way to guarantee your UI handles every edge case before you even run it.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ chupathingyce/chupathingyce
Play Halo: Combat Evolved Online on Your Phone and PC
Bring the original Xbox Halo back to life on your phone, laptop, or desktop with ChupathingyCE. This community project lets you run the classic shooter natively, meaning no heavy emulator needed. The coolest part is the built-in online system that connects you with friends directly, skipping the annoying port forwarding setup. You can even host dedicated servers to keep the action going around the clock. It is a great way to experience the 2001 classic with modern quality and smooth networking. Grab a controller and see what the community has built for this timeless title.
π @hackernewsgithubprojects
Play Halo: Combat Evolved Online on Your Phone and PC
Bring the original Xbox Halo back to life on your phone, laptop, or desktop with ChupathingyCE. This community project lets you run the classic shooter natively, meaning no heavy emulator needed. The coolest part is the built-in online system that connects you with friends directly, skipping the annoying port forwarding setup. You can even host dedicated servers to keep the action going around the clock. It is a great way to experience the 2001 classic with modern quality and smooth networking. Grab a controller and see what the community has built for this timeless title.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ austin-starks/public-portfolio-challenge
Real Money, AI Traders: Public Portfolio Challenge
Put twenty-five thousand dollars on the line and let four different A.I. agents build a trading strategy, with every single decision and failure shown in public. This is the Public Portfolio Challenge, an experiment that proves artificial intelligence can handle real brokerage accounts without hiding behind backtest screenshots. The most surprising part is the honesty: the project logs every bug, panic sell, and rejected model in a shared repository, so anyone can see exactly how the software thinks. You get to watch live performance data update in real time, proving that the results are genuine and not just a clever trick.
π @hackernewsgithubprojects
Real Money, AI Traders: Public Portfolio Challenge
Put twenty-five thousand dollars on the line and let four different A.I. agents build a trading strategy, with every single decision and failure shown in public. This is the Public Portfolio Challenge, an experiment that proves artificial intelligence can handle real brokerage accounts without hiding behind backtest screenshots. The most surprising part is the honesty: the project logs every bug, panic sell, and rejected model in a shared repository, so anyone can see exactly how the software thinks. You get to watch live performance data update in real time, proving that the results are genuine and not just a clever trick.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ aqm857886159/nomi
Nomi: The AI Video Studio That Doesn't Own You
Nomi is the open-source video studio that lets your AI agent actually cut the film. It isn't just another chatbot wrapper; it is a real editor where you plug in your own models. The agent takes your idea, breaks it into shots, generates the visuals, and places them on a timeline you can scrub and fix. The coolest part? You pay the model providers directly, so there is no middleman taking a cut. Your prompts, keys, and projects stay on your machine, which means total privacy. No accounts, no hidden telemetry, just a clean, local-first workflow that puts you in the director's chair.
π @hackernewsgithubprojects
Nomi: The AI Video Studio That Doesn't Own You
Nomi is the open-source video studio that lets your AI agent actually cut the film. It isn't just another chatbot wrapper; it is a real editor where you plug in your own models. The agent takes your idea, breaks it into shots, generates the visuals, and places them on a timeline you can scrub and fix. The coolest part? You pay the model providers directly, so there is no middleman taking a cut. Your prompts, keys, and projects stay on your machine, which means total privacy. No accounts, no hidden telemetry, just a clean, local-first workflow that puts you in the director's chair.
π @hackernewsgithubprojects