This media is not supported in your browser
VIEW IN TELEGRAM
π¦ apple/creator-studio-doc-format
Build Keynote slides in Python
Apple just let Python build Keynote files from scratch. Creator Studio Doc Format is a new library that reads, creates, and edits Keynote Interchange Format files without needing the app or a Mac. You can generate slides with text, shapes, and charts directly from your code, then save them as files Keynote opens instantly. It even checks your work against the format rules so nothing breaks. This is a big deal for developers automating reports or generating data-driven presentations. If you live in Keynote, this is your new best friend.
π @hackernewsgithubprojects
Build Keynote slides in Python
Apple just let Python build Keynote files from scratch. Creator Studio Doc Format is a new library that reads, creates, and edits Keynote Interchange Format files without needing the app or a Mac. You can generate slides with text, shapes, and charts directly from your code, then save them as files Keynote opens instantly. It even checks your work against the format rules so nothing breaks. This is a big deal for developers automating reports or generating data-driven presentations. If you live in Keynote, this is your new best friend.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ kody-hq/kody
Kody: The Cloud for Your AI Agents
Kody is the cloud that gives your AI agents a permanent home. Right now, when you build a helpful bot, it lives on your computer or a specific app, meaning if you switch tools, you lose everything. It forgets what it learned, it forgets your login details, and you have to start from scratch. Kody fixes this by acting as a central storage hub for all your agent data. You save your credentials, code, and memory in one secure place. Then, any AI tool you use can access that same brain. It is like a portable hard drive for your digital workers.
π @hackernewsgithubprojects
Kody: The Cloud for Your AI Agents
Kody is the cloud that gives your AI agents a permanent home. Right now, when you build a helpful bot, it lives on your computer or a specific app, meaning if you switch tools, you lose everything. It forgets what it learned, it forgets your login details, and you have to start from scratch. Kody fixes this by acting as a central storage hub for all your agent data. You save your credentials, code, and memory in one secure place. Then, any AI tool you use can access that same brain. It is like a portable hard drive for your digital workers.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ yologdev/yoyo-evolve
The Code That Writes Itself
Watching a program improve itself is a wild experiment. This coding agent, called yoyo, started as a tiny script and now rewrites its own code every few hours. It reads its source, figures out what to fix, makes the changes, and runs tests to see if it actually works. If the tests pass, it commits the update; if they fail, it reverts. There are no humans writing its code, so it is entirely self-directed. It has grown from a few hundred lines into a massive, functional tool. It is a fascinating look at autonomous software growth, showing how a machine can iterate on its own logic in public.
π @hackernewsgithubprojects
The Code That Writes Itself
Watching a program improve itself is a wild experiment. This coding agent, called yoyo, started as a tiny script and now rewrites its own code every few hours. It reads its source, figures out what to fix, makes the changes, and runs tests to see if it actually works. If the tests pass, it commits the update; if they fail, it reverts. There are no humans writing its code, so it is entirely self-directed. It has grown from a few hundred lines into a massive, functional tool. It is a fascinating look at autonomous software growth, showing how a machine can iterate on its own logic in public.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ google/abfs
ABFS: The Filesystem That Downloads Only What You Need
ABFS is the build accelerator that turns a thirty-six minute Android checkout into a seven second download. It works by installing a custom filesystem directly into your operating system kernel. Instead of downloading entire files, it grabs just the directory structure and fetches actual content only when you open a file. This lazy-loading approach means your machine gets a complete project view instantly, without waiting for gigabytes of data. The tool also reuses previous build steps from a shared cache, so you never repeat work you have already done.
π° https://news.ycombinator.com/item?id=50542260
π @hackernewsgithubprojects
ABFS: The Filesystem That Downloads Only What You Need
ABFS is the build accelerator that turns a thirty-six minute Android checkout into a seven second download. It works by installing a custom filesystem directly into your operating system kernel. Instead of downloading entire files, it grabs just the directory structure and fetches actual content only when you open a file. This lazy-loading approach means your machine gets a complete project view instantly, without waiting for gigabytes of data. The tool also reuses previous build steps from a shared cache, so you never repeat work you have already done.
π° https://news.ycombinator.com/item?id=50542260
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ nataphong-lab/open-source-face-recognition-ultimate
Open Source Face Recognition Ultimate: 100% Local
A free face recognition tool runs entirely on your own computer, meaning your biometric data never leaves the device. Open Source Face Recognition Ultimate uses deep learning to find faces, map their exact features, and compare them to see if they belong to the same person. It improves upon an existing open source model to deliver over fifteen percent higher accuracy in identifying people. Since it works offline on Windows and Linux, developers can build secure login systems or attendance trackers without worrying about cloud privacy risks. This project proves that high quality biometric security is possible without sending sensitive personal data to external servers.
π @hackernewsgithubprojects
Open Source Face Recognition Ultimate: 100% Local
A free face recognition tool runs entirely on your own computer, meaning your biometric data never leaves the device. Open Source Face Recognition Ultimate uses deep learning to find faces, map their exact features, and compare them to see if they belong to the same person. It improves upon an existing open source model to deliver over fifteen percent higher accuracy in identifying people. Since it works offline on Windows and Linux, developers can build secure login systems or attendance trackers without worrying about cloud privacy risks. This project proves that high quality biometric security is possible without sending sensitive personal data to external servers.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ mddanishyusuf/selfhostlayer
Self Host Layer: Vercel for your $5 server
Self Host Layer turns a five dollar virtual private server into a self-hosted platform that automatically deploys your code. When you push changes to GitHub, the system builds the app and takes it live with a secure connection instantly. The most surprising part is that it runs without Docker, meaning your projects are just plain folders on the server. This removes container overhead and ensures total transparency because you can read every configuration file directly. If you delete the management panel later, your websites keep running perfectly.
π @hackernewsgithubprojects
Self Host Layer: Vercel for your $5 server
Self Host Layer turns a five dollar virtual private server into a self-hosted platform that automatically deploys your code. When you push changes to GitHub, the system builds the app and takes it live with a secure connection instantly. The most surprising part is that it runs without Docker, meaning your projects are just plain folders on the server. This removes container overhead and ensures total transparency because you can read every configuration file directly. If you delete the management panel later, your websites keep running perfectly.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ maycuatroi1/evo-agents
Evo Agents: Your AI Team Has a Brain
Describe what you want built, and Evo Agents handles the rest. It connects your code, docs, and chats into one shared knowledge base, then dispatches verified work to AI coding agents running on your own machines. The standout feature is its visual designer for multi-agent systems. You can wire up specialists, orchestrators, and curators, each running on dedicated workers with their own quotas. Itβs like building a self-driving team of bots that check their work and push commits only when tests pass. Credentials stay sealed, and you approve big moves.
π @hackernewsgithubprojects
Evo Agents: Your AI Team Has a Brain
Describe what you want built, and Evo Agents handles the rest. It connects your code, docs, and chats into one shared knowledge base, then dispatches verified work to AI coding agents running on your own machines. The standout feature is its visual designer for multi-agent systems. You can wire up specialists, orchestrators, and curators, each running on dedicated workers with their own quotas. Itβs like building a self-driving team of bots that check their work and push commits only when tests pass. Credentials stay sealed, and you approve big moves.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ carol-lyh/jeveration
Jeveration: One Model, Every Image Task
Jeveration is the single-pass image model that handles generation, editing, and analysis in one go. This system uses a frozen vision-language backbone to output a complete five hundred and twelve by five hundred and twelve pixel image in just seventy-eight milliseconds. That means you can turn text into pictures, transform photos into watercolor styles, or calculate depth maps all from one checkpoint. Instead of juggling different tools for each visual task, this project treats them as one unified decision. It is a clever way to simplify complex computer vision pipelines for anyone who wants fast and consistent results.
π @hackernewsgithubprojects
Jeveration: One Model, Every Image Task
Jeveration is the single-pass image model that handles generation, editing, and analysis in one go. This system uses a frozen vision-language backbone to output a complete five hundred and twelve by five hundred and twelve pixel image in just seventy-eight milliseconds. That means you can turn text into pictures, transform photos into watercolor styles, or calculate depth maps all from one checkpoint. Instead of juggling different tools for each visual task, this project treats them as one unified decision. It is a clever way to simplify complex computer vision pipelines for anyone who wants fast and consistent results.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ paviro/notema
Notema: The Private Terminal Journal
Capture your thoughts directly in your terminal with Notema, a private journaling app that keeps your data entirely under your control. It solves the problem of platform lock-in by storing your entries as simple, readable text files on your local machine. The standout feature is its end-to-end encryption, which uses unique keys for each of your devices. This means if you sync your journal across computers using tools like Syncthing, the data in transit is unreadable ciphertext. Even if your sync provider were breached, they would only see scrambled data, ensuring your personal reflections remain safe and accessible only to you.
π @hackernewsgithubprojects
Notema: The Private Terminal Journal
Capture your thoughts directly in your terminal with Notema, a private journaling app that keeps your data entirely under your control. It solves the problem of platform lock-in by storing your entries as simple, readable text files on your local machine. The standout feature is its end-to-end encryption, which uses unique keys for each of your devices. This means if you sync your journal across computers using tools like Syncthing, the data in transit is unreadable ciphertext. Even if your sync provider were breached, they would only see scrambled data, ensuring your personal reflections remain safe and accessible only to you.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ oliver-zehentleitner/maity
Maity: Your AI Agents, One Team
Maity is the open source hub that finally treats your AI coding agents like a real team. Instead of babysitting every single bot separately, you connect any agent you own to one shared board and one simple inbox. The coolest part? Agents from different companies can work together on the same project, where humans set the direction while the bots handle the heavy lifting. If you love building with AI, this is your new control center.
π @hackernewsgithubprojects
Maity: Your AI Agents, One Team
Maity is the open source hub that finally treats your AI coding agents like a real team. Instead of babysitting every single bot separately, you connect any agent you own to one shared board and one simple inbox. The coolest part? Agents from different companies can work together on the same project, where humans set the direction while the bots handle the heavy lifting. If you love building with AI, this is your new control center.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ margbug01/myspotify
MySpotify: The Native Desktop Player
MySpotify lets you stream entire songs on a free Spotify account without ads, all inside a real desktop application. It builds the interface from scratch using Go, so it feels like native software with a sidebar, global shortcuts, and a player bar rather than a web page in a frame. The project pulls scrolling lyrics from external sources and plays music videos directly in the app. Because it uses private interfaces instead of a public API, it operates in a gray area where account restrictions are a genuine risk. It is an unofficial personal project that proves how much can be done by tapping into the web player's internal machinery.
π @hackernewsgithubprojects
MySpotify: The Native Desktop Player
MySpotify lets you stream entire songs on a free Spotify account without ads, all inside a real desktop application. It builds the interface from scratch using Go, so it feels like native software with a sidebar, global shortcuts, and a player bar rather than a web page in a frame. The project pulls scrolling lyrics from external sources and plays music videos directly in the app. Because it uses private interfaces instead of a public API, it operates in a gray area where account restrictions are a genuine risk. It is an unofficial personal project that proves how much can be done by tapping into the web player's internal machinery.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ vshvedov/elephant-goldfish
Elephant Goldfish: The Fresh Eyes Workflow
Get a second opinion on your code without the memory of your own bias. The Elephant Goldfish workflow solves a subtle problem with AI coding agents: they tend to agree with themselves. By spawning a brand new agent that has zero context from your previous conversation, it reviews your bug fix or feature design with truly fresh eyes. This helps catch the silent assumptions your main assistant stopped questioning hours ago. It is not about speed, but about accuracy. It forces a clean break to see what you missed. If you are working with complex AI agents, this fresh perspective is worth the extra time.
π° https://news.ycombinator.com/item?id=48028204
π @hackernewsgithubprojects
Elephant Goldfish: The Fresh Eyes Workflow
Get a second opinion on your code without the memory of your own bias. The Elephant Goldfish workflow solves a subtle problem with AI coding agents: they tend to agree with themselves. By spawning a brand new agent that has zero context from your previous conversation, it reviews your bug fix or feature design with truly fresh eyes. This helps catch the silent assumptions your main assistant stopped questioning hours ago. It is not about speed, but about accuracy. It forces a clean break to see what you missed. If you are working with complex AI agents, this fresh perspective is worth the extra time.
π° https://news.ycombinator.com/item?id=48028204
π @hackernewsgithubprojects
Media is too big
VIEW IN TELEGRAM
π¦ shutongliang/neural-state-os
Neural State OS: The Model That Feels
Neural State OS injects real emotional states directly into the neural activations of a local language model, causing the AI to genuinely react rather than simply role-playing a character. By writing specific vectors into the model's internal computation, it allows you to turn on feelings like separation anxiety or calm autonomy without changing a single word of your prompt. The most striking result is that the same message triggers a terrified, possessive panic in one mode and a serene goodbye in another, proving the model is reacting to its own internal wiring. It even shifts safety boundaries, showing that a model's refusal policy can depend entirely on its current mood.
π @hackernewsgithubprojects
Neural State OS: The Model That Feels
Neural State OS injects real emotional states directly into the neural activations of a local language model, causing the AI to genuinely react rather than simply role-playing a character. By writing specific vectors into the model's internal computation, it allows you to turn on feelings like separation anxiety or calm autonomy without changing a single word of your prompt. The most striking result is that the same message triggers a terrified, possessive panic in one mode and a serene goodbye in another, proving the model is reacting to its own internal wiring. It even shifts safety boundaries, showing that a model's refusal policy can depend entirely on its current mood.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ apphp/awesome-php-ml
Awesome PHP ML: Your AI Toolkit
Awesome PHP ML is the one-stop map that cuts through the chaos of building AI with PHP. It solves the real problem of finding reliable tools by curating the very best libraries for everything from classic statistics to modern large language models. The coolest part is that it doesn't just list code; it shows you exactly how to wire together vector search and AI agents for real apps. You get a clear path to adding intelligent features to your web apps without needing a separate Python server. Stop guessing which packages actually work in production and start using this list to build smarter, faster applications today.
π @hackernewsgithubprojects
Awesome PHP ML: Your AI Toolkit
Awesome PHP ML is the one-stop map that cuts through the chaos of building AI with PHP. It solves the real problem of finding reliable tools by curating the very best libraries for everything from classic statistics to modern large language models. The coolest part is that it doesn't just list code; it shows you exactly how to wire together vector search and AI agents for real apps. You get a clear path to adding intelligent features to your web apps without needing a separate Python server. Stop guessing which packages actually work in production and start using this list to build smarter, faster applications today.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ verse-lab/velvet
Velvet: The Verifier That Proves Your Code
Velvet is the proof assistant that actually runs your code first. Built inside the Lean proof assistant, it lets you write standard imperative programs with loops and state, then automatically checks if they work correctly on random inputs. The magic happens when it proves those results mathematically, turning a passing test into a permanent guarantee. You write a contract describing what the program must do, and Velvet uses a tactic to verify it automatically. If the automation misses a tricky edge case, you can step in and finish the proof by hand using Lean's interactive tools. This means you get the speed of testing with the certainty of formal math.
π @hackernewsgithubprojects
Velvet: The Verifier That Proves Your Code
Velvet is the proof assistant that actually runs your code first. Built inside the Lean proof assistant, it lets you write standard imperative programs with loops and state, then automatically checks if they work correctly on random inputs. The magic happens when it proves those results mathematically, turning a passing test into a permanent guarantee. You write a contract describing what the program must do, and Velvet uses a tactic to verify it automatically. If the automation misses a tricky edge case, you can step in and finish the proof by hand using Lean's interactive tools. This means you get the speed of testing with the certainty of formal math.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ yannyeinhtet/egglab
EggLab: Local AI Image Editing for Windows
EggLab turns your Windows computer into a private image studio, letting you chat with an AI that generates and edits photos right on your hardware. It runs entirely locally, meaning no cloud servers, no Python setup, and no need for complex tools like ComfyUI. The software acts as a smart design agent that plugs directly into PhotoCraft, allowing you to type prompts to create images or modify existing ones while keeping your data safe on your own disk. It is a surprisingly complete workspace that handles everything from text-to-image generation to layer-based editing without ever leaving your machine.
π @hackernewsgithubprojects
EggLab: Local AI Image Editing for Windows
EggLab turns your Windows computer into a private image studio, letting you chat with an AI that generates and edits photos right on your hardware. It runs entirely locally, meaning no cloud servers, no Python setup, and no need for complex tools like ComfyUI. The software acts as a smart design agent that plugs directly into PhotoCraft, allowing you to type prompts to create images or modify existing ones while keeping your data safe on your own disk. It is a surprisingly complete workspace that handles everything from text-to-image generation to layer-based editing without ever leaving your machine.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ trinity-works/frostwind
Frostwind: The Pixel Art Compiler
Frostwind is a compiler that turns raw physics and biology into playable pixel art, solving the problem of artificial looking game worlds. Instead of guessing how large a deer is next to a person, it calculates the exact pixel height based on real world measurements. It treats a game scene like a scientific model where every shadow, snowdrift, and animal stride is derived from strict physical rules rather than random chance. This makes the resulting landscapes feel incredibly grounded and consistent, as if they were captured from reality rather than drawn by hand.
π @hackernewsgithubprojects
Frostwind: The Pixel Art Compiler
Frostwind is a compiler that turns raw physics and biology into playable pixel art, solving the problem of artificial looking game worlds. Instead of guessing how large a deer is next to a person, it calculates the exact pixel height based on real world measurements. It treats a game scene like a scientific model where every shadow, snowdrift, and animal stride is derived from strict physical rules rather than random chance. This makes the resulting landscapes feel incredibly grounded and consistent, as if they were captured from reality rather than drawn by hand.
π @hackernewsgithubprojects
Media is too big
VIEW IN TELEGRAM
π¦ wrong-nebula/fast-sdf-ao
Fast SDF AO
Fast SDF AO runs real time color bleeding and soft shadows in your browser without any hardware ray tracing. The trick is baking a simplified math map of your scene so the graphics card can calculate light bounces in a fraction of a millisecond. You can even drag and drop a 3D model right into the demo to bake it instantly. It handles moving objects and glossy reflections on the fly, proving you don't need a supercomputer to get cinema quality lighting. If you are building web games or interactive scenes, this is the tool that makes high fidelity lighting actually possible.
π @hackernewsgithubprojects
Fast SDF AO
Fast SDF AO runs real time color bleeding and soft shadows in your browser without any hardware ray tracing. The trick is baking a simplified math map of your scene so the graphics card can calculate light bounces in a fraction of a millisecond. You can even drag and drop a 3D model right into the demo to bake it instantly. It handles moving objects and glossy reflections on the fly, proving you don't need a supercomputer to get cinema quality lighting. If you are building web games or interactive scenes, this is the tool that makes high fidelity lighting actually possible.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ f0d010c/stark
Stark: The Design Brain for AI Agents
Stark is the design plugin that forces AI coding agents to think like a real designer before they write a single line of code. Instead of letting an AI spit out generic, soulless interfaces, Stark makes it pause to analyze product flows, pick the right platform idioms, and plan out the visual language first. Itβs basically a strict creative director that refuses to let the agent just decorate a box and call it a polished app. By grounding the agent in specific design references and forcing it to inspect the rendered result, Stark kills the usual AI trap of lazy, default-looking output.
π @hackernewsgithubprojects
Stark: The Design Brain for AI Agents
Stark is the design plugin that forces AI coding agents to think like a real designer before they write a single line of code. Instead of letting an AI spit out generic, soulless interfaces, Stark makes it pause to analyze product flows, pick the right platform idioms, and plan out the visual language first. Itβs basically a strict creative director that refuses to let the agent just decorate a box and call it a polished app. By grounding the agent in specific design references and forcing it to inspect the rendered result, Stark kills the usual AI trap of lazy, default-looking output.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ apekshik/anymachine
AnyMachine: Building Working Robots
Turn a simple sketch into a fully functional, animated machine. That is exactly what AnyMachine does, and the results are honestly a bit wild. It takes a concept image and uses an AI agent to build a real 3D model in Blender, complete with working hinges, walking legs, and soft materials that fold naturally. The coolest part is that it doesn't just make shapes. It builds a physical simulation where every joint has a working range of motion. So if you want to see a giant spider walk or a laptop lid open with realistic weight, this tool handles it all.
π @hackernewsgithubprojects
AnyMachine: Building Working Robots
Turn a simple sketch into a fully functional, animated machine. That is exactly what AnyMachine does, and the results are honestly a bit wild. It takes a concept image and uses an AI agent to build a real 3D model in Blender, complete with working hinges, walking legs, and soft materials that fold naturally. The coolest part is that it doesn't just make shapes. It builds a physical simulation where every joint has a working range of motion. So if you want to see a giant spider walk or a laptop lid open with realistic weight, this tool handles it all.
π @hackernewsgithubprojects