This media is not supported in your browser
VIEW IN TELEGRAM
π¦ jumbomuffin777/frontier-ui
Frontier UI: ChatGPT's interface for open models
Your local AI chat just started drawing real street maps with walking routes and pulling live photos directly into the conversation. Frontier UI is a single file you drop into Open WebUI that turns plain text answers into interactive cards. It finds real images from Wikipedia, verifies locations on a live map, and even builds downloadable Excel files from your chat. No API keys or complex setup needed. It basically gives any open model the same rich visual interface people expect from ChatGPT. If you run local AI, this is the upgrade you need.
π @hackernewsgithubprojects
Frontier UI: ChatGPT's interface for open models
Your local AI chat just started drawing real street maps with walking routes and pulling live photos directly into the conversation. Frontier UI is a single file you drop into Open WebUI that turns plain text answers into interactive cards. It finds real images from Wikipedia, verifies locations on a live map, and even builds downloadable Excel files from your chat. No API keys or complex setup needed. It basically gives any open model the same rich visual interface people expect from ChatGPT. If you run local AI, this is the upgrade you need.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ grimmory-tools/grimmory
Grimmory: Your Home Library, Upgraded
Build your own private library that actually works. Grimmory is a self-hosted system for storing ebooks, comics, and audiobooks, letting you read them right in your browser while automatically pulling in cover art and reviews. The coolest part is the BookDrop feature, where you just drop files into a folder and it detects, enriches, and queues them for you. It is perfect for anyone who wants full control over their digital collection without relying on big tech servers. If you love reading, this is the setup you need to stop juggling files and start enjoying the books. Go grab your stack and get started.
π @hackernewsgithubprojects
Grimmory: Your Home Library, Upgraded
Build your own private library that actually works. Grimmory is a self-hosted system for storing ebooks, comics, and audiobooks, letting you read them right in your browser while automatically pulling in cover art and reviews. The coolest part is the BookDrop feature, where you just drop files into a folder and it detects, enriches, and queues them for you. It is perfect for anyone who wants full control over their digital collection without relying on big tech servers. If you love reading, this is the setup you need to stop juggling files and start enjoying the books. Go grab your stack and get started.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ serioushornet/koreader.patches
Koreader Patches: A Visual Overhaul for E-Readers
Your bookshelf on an e-reader can finally look as polished as a physical display case, thanks to a collection of visual patches for Koreader. Instead of a plain grid, this suite gives every cover rounded corners and stretches them to fit their spaces perfectly, creating a uniform and tidy mosaic. It strips away the default, often cluttered status bars and replaces them with clean, minimalist badges that show your reading progress and collection status without covering up the artwork. If you have ever felt that the default interface on your device looked too plain or busy, this is a simple way to make your library feel curated and visually cohesive.
π @hackernewsgithubprojects
Koreader Patches: A Visual Overhaul for E-Readers
Your bookshelf on an e-reader can finally look as polished as a physical display case, thanks to a collection of visual patches for Koreader. Instead of a plain grid, this suite gives every cover rounded corners and stretches them to fit their spaces perfectly, creating a uniform and tidy mosaic. It strips away the default, often cluttered status bars and replaces them with clean, minimalist badges that show your reading progress and collection status without covering up the artwork. If you have ever felt that the default interface on your device looked too plain or busy, this is a simple way to make your library feel curated and visually cohesive.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ crmne/zapfast
ZapFast: A Native WhatsApp Client
A desktop WhatsApp client that opens in under one second without a browser engine is a technical feat called ZapFast. Built with Rust, it acts as a native companion to your phone rather than a web page loaded into a browser, delivering a lightweight experience for Linux, macOS, and Windows. This approach solves the problem of slow startup times and heavy memory usage associated with traditional web-based chat clients. The result is a snappy and efficient way to manage your messages on a desktop. The key takeaway is that native applications can deliver the speed and performance you expect from modern software.
π @hackernewsgithubprojects
ZapFast: A Native WhatsApp Client
A desktop WhatsApp client that opens in under one second without a browser engine is a technical feat called ZapFast. Built with Rust, it acts as a native companion to your phone rather than a web page loaded into a browser, delivering a lightweight experience for Linux, macOS, and Windows. This approach solves the problem of slow startup times and heavy memory usage associated with traditional web-based chat clients. The result is a snappy and efficient way to manage your messages on a desktop. The key takeaway is that native applications can deliver the speed and performance you expect from modern software.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ aa1143/namiagent
NamiAgent: See Inside the AI Agent
NamiAgent is the educational agent that lets you see every single thought a machine makes. Most AI tools feel like magic boxes, but this project strips away the mystery to show you the actual loop. It starts with just a hundred lines of code and builds up skills one by one. The coolest part is a simple web page that displays exactly what the AI is thinking and what files it touches in real time. You stop guessing how these tools work and start understanding the engine behind them. This project turns a confusing black box into a clear, transparent lesson.
π @hackernewsgithubprojects
NamiAgent: See Inside the AI Agent
NamiAgent is the educational agent that lets you see every single thought a machine makes. Most AI tools feel like magic boxes, but this project strips away the mystery to show you the actual loop. It starts with just a hundred lines of code and builds up skills one by one. The coolest part is a simple web page that displays exactly what the AI is thinking and what files it touches in real time. You stop guessing how these tools work and start understanding the engine behind them. This project turns a confusing black box into a clear, transparent lesson.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ microsoft/sdbench
SDBench: The Medical AI Test That Actually Works
SDBench is the diagnostic benchmark that finally measures how well AI thinks through medical cases one step at a time. Instead of just asking a model to guess a disease from a static description, this tool lets the AI act like a real doctor. It can ask follow-up questions and order specific lab tests, then wait for the answers before deciding on a diagnosis. This setup, published in Science, reveals whether a model can truly reason through complex clinical scenarios rather than just pattern-matching. It even tracks the cost of every test ordered, giving you a clear picture of efficiency.
π @hackernewsgithubprojects
SDBench: The Medical AI Test That Actually Works
SDBench is the diagnostic benchmark that finally measures how well AI thinks through medical cases one step at a time. Instead of just asking a model to guess a disease from a static description, this tool lets the AI act like a real doctor. It can ask follow-up questions and order specific lab tests, then wait for the answers before deciding on a diagnosis. This setup, published in Science, reveals whether a model can truly reason through complex clinical scenarios rather than just pattern-matching. It even tracks the cost of every test ordered, giving you a clear picture of efficiency.
π @hackernewsgithubprojects
Media is too big
VIEW IN TELEGRAM
π¦ apexin-ai/apex-35b-v1.0
Apex-35B: The Self-Improving AI
Apex-35B is the AI model that actually improves itself while it works. Most software just sits there, but this one uses an automated system to find its own weak spots and generate new training data to fix them. It even optimizes its own internal code to run faster on your hardware. It is built on a foundation that lets it handle massive amounts of text at once, which is why it shines at complex scientific and coding tasks. You do not have to manage its thinking; it decides for itself how long to reason on each step. The result is a tool that gets better the more you use it.
π @hackernewsgithubprojects
Apex-35B: The Self-Improving AI
Apex-35B is the AI model that actually improves itself while it works. Most software just sits there, but this one uses an automated system to find its own weak spots and generate new training data to fix them. It even optimizes its own internal code to run faster on your hardware. It is built on a foundation that lets it handle massive amounts of text at once, which is why it shines at complex scientific and coding tasks. You do not have to manage its thinking; it decides for itself how long to reason on each step. The result is a tool that gets better the more you use it.
π @hackernewsgithubprojects
Media is too big
VIEW IN TELEGRAM
π¦ t8y2/motty
Motty: One Window for All Remote Tools
Stop juggling ten different apps just to manage your servers. Motty is a native desktop application that bundles secure terminal access, file transfers, and remote desktop connections into a single, lightweight window. Built with Go, it completely avoids the heavy, battery-draining web browser tech that most modern apps rely on. The real surprise here is the built-in AI assistant. It sits right alongside your terminal and file manager, ready to help you write complex commands or troubleshoot errors instantly. You get a powerful, focused workspace for your entire remote workflow without the clutter of separate tools. Itβs a clean, fast way to handle all your server tasks in one place.
π @hackernewsgithubprojects
Motty: One Window for All Remote Tools
Stop juggling ten different apps just to manage your servers. Motty is a native desktop application that bundles secure terminal access, file transfers, and remote desktop connections into a single, lightweight window. Built with Go, it completely avoids the heavy, battery-draining web browser tech that most modern apps rely on. The real surprise here is the built-in AI assistant. It sits right alongside your terminal and file manager, ready to help you write complex commands or troubleshoot errors instantly. You get a powerful, focused workspace for your entire remote workflow without the clutter of separate tools. Itβs a clean, fast way to handle all your server tasks in one place.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ prefrontal-ai/latent-sparse-attention
Latent Sparse Attention: 72x Faster Long Context AI
Latent Sparse Attention makes large language models process ten million token contexts up to seventy-two times faster than standard attention. This is not a new math formula, but a smarter way to look. Instead of checking every single word against every other word, the system groups similar ideas into clusters first. It then only reads the most relevant clusters, skipping the rest. This means the model skips the boring parts and focuses on what matters. You get near-perfect accuracy while running massively faster. It works on existing models without retraining them. It is a practical fix for the slowness that kills long context usage.
π @hackernewsgithubprojects
Latent Sparse Attention: 72x Faster Long Context AI
Latent Sparse Attention makes large language models process ten million token contexts up to seventy-two times faster than standard attention. This is not a new math formula, but a smarter way to look. Instead of checking every single word against every other word, the system groups similar ideas into clusters first. It then only reads the most relevant clusters, skipping the rest. This means the model skips the boring parts and focuses on what matters. You get near-perfect accuracy while running massively faster. It works on existing models without retraining them. It is a practical fix for the slowness that kills long context usage.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ samuelcorsan/reforma-digital
Reforma Digital: Make Government Websites Actually Make Sense
Take a breath and imagine finally understanding a government website without needing a law degree. This community project called Reforma Digital overlays helpful, plain-language guides onto official Spanish administrative sites, like your tax office or ID renewal portal. It does not change the actual rules or bypass security checks; it simply helps you find the right form and understand what every button does. It is an open-source effort to make bureaucracy less confusing, and the most exciting part is that it works locally in your browser. The code is public, so anyone can see exactly how it helps and even suggest improvements.
π @hackernewsgithubprojects
Reforma Digital: Make Government Websites Actually Make Sense
Take a breath and imagine finally understanding a government website without needing a law degree. This community project called Reforma Digital overlays helpful, plain-language guides onto official Spanish administrative sites, like your tax office or ID renewal portal. It does not change the actual rules or bypass security checks; it simply helps you find the right form and understand what every button does. It is an open-source effort to make bureaucracy less confusing, and the most exciting part is that it works locally in your browser. The code is public, so anyone can see exactly how it helps and even suggest improvements.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ arvarik/contrack
Contrack: Your Contacts Finally Live Here
Consolidate every person in your life into one private, self-hosted hub. Contrack pulls contacts from Apple, Google, LinkedIn, and Facebook, merging duplicates so you get one clean profile per person. The coolest part is the enrichment step: you paste a quick note about a meeting, and it automatically extracts dates, tags, and reminders into a full profile. It even fills in missing web data with cited sources. Plus, it tracks who needs a follow-up so relationships never drift. It is your personal network, finally owned and organized. Check out Contrack to stop losing track of the people you care about.
π° https://news.ycombinator.com/item?id=50169363
π @hackernewsgithubprojects
Contrack: Your Contacts Finally Live Here
Consolidate every person in your life into one private, self-hosted hub. Contrack pulls contacts from Apple, Google, LinkedIn, and Facebook, merging duplicates so you get one clean profile per person. The coolest part is the enrichment step: you paste a quick note about a meeting, and it automatically extracts dates, tags, and reminders into a full profile. It even fills in missing web data with cited sources. Plus, it tracks who needs a follow-up so relationships never drift. It is your personal network, finally owned and organized. Check out Contrack to stop losing track of the people you care about.
π° https://news.ycombinator.com/item?id=50169363
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ athola/claude-night-market
Claude Night Market: Your AI Coding Assistant Has a Grocery Store
A giant marketplace just dropped for AI coding assistants. It lets you pick and choose exactly which tools to install for your workflow. Think of it as a curated grocery store for your computer brain. You grab the plugin for code reviews, or the one that forces you to write tests before code. This solves the problem of your AI getting too smart or too lazy. You control its behavior with specific rules. It is useful because you stop wrestling with generic AI and start directing a precise team of digital workers. It is a practical way to make your AI assistant actually useful for real software projects.
π @hackernewsgithubprojects
Claude Night Market: Your AI Coding Assistant Has a Grocery Store
A giant marketplace just dropped for AI coding assistants. It lets you pick and choose exactly which tools to install for your workflow. Think of it as a curated grocery store for your computer brain. You grab the plugin for code reviews, or the one that forces you to write tests before code. This solves the problem of your AI getting too smart or too lazy. You control its behavior with specific rules. It is useful because you stop wrestling with generic AI and start directing a precise team of digital workers. It is a practical way to make your AI assistant actually useful for real software projects.
π @hackernewsgithubprojects
β€1
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ meshy-dev/meshy-mcp-server
Meshy MCP Server
Turn text or photos into printable, animated 3D models directly from your AI chat. The Meshy MCP Server bridges your favorite coding assistants with the Meshy platform, allowing you to generate characters, rig them, and even check if they will fit on a 3D printer without ever leaving your terminal. Instead of opening a separate website and clicking through menus, you simply describe what you need in plain language. The agent handles the complex steps, from generating the mesh to applying high resolution textures, so you can focus on the creative idea rather than the technical plumbing.
π @hackernewsgithubprojects
Meshy MCP Server
Turn text or photos into printable, animated 3D models directly from your AI chat. The Meshy MCP Server bridges your favorite coding assistants with the Meshy platform, allowing you to generate characters, rig them, and even check if they will fit on a 3D printer without ever leaving your terminal. Instead of opening a separate website and clicking through menus, you simply describe what you need in plain language. The agent handles the complex steps, from generating the mesh to applying high resolution textures, so you can focus on the creative idea rather than the technical plumbing.
π @hackernewsgithubprojects
Media is too big
VIEW IN TELEGRAM
π¦ drawthingsai/local-code-community
Local Code: Your Private AI Agent
Local Code turns your Mac, iPad, or iPhone into a private workhorse that actually gets things done. Instead of just chatting, it runs a massive DeepSeek model entirely on your device, meaning your data never leaves your hands. It comes with a built-in sandbox containing Python, Node, and a shell, so it can analyze data, build charts, and even edit documents without touching your main files. The coolest part? It handles models bigger than your RAM by streaming them from disk, so you get serious brainpower without the cloud overhead. If you want an AI assistant that respects your privacy and actually builds stuff, this is your new secret weapon.
π° https://news.ycombinator.com/item?id=50281590
π @hackernewsgithubprojects
Local Code: Your Private AI Agent
Local Code turns your Mac, iPad, or iPhone into a private workhorse that actually gets things done. Instead of just chatting, it runs a massive DeepSeek model entirely on your device, meaning your data never leaves your hands. It comes with a built-in sandbox containing Python, Node, and a shell, so it can analyze data, build charts, and even edit documents without touching your main files. The coolest part? It handles models bigger than your RAM by streaming them from disk, so you get serious brainpower without the cloud overhead. If you want an AI assistant that respects your privacy and actually builds stuff, this is your new secret weapon.
π° https://news.ycombinator.com/item?id=50281590
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ mrlong0129/yominpost
YominPost
Turn a single URL, text snippet, or image into custom, platform-native social media posts without relying on third-party cloud tools. YominPost runs directly on your own machine, automatically picking the best post format for channels like X or LinkedIn, scoring its own copy, and running pre-publish risk checks to block prompt leaks or engagement bait before anything goes live. It hooks up to coding agents like Claude Code or Cursor through a local tool integration, letting AI assist with drafting while leaving you in full control of final approval.
π @hackernewsgithubprojects
YominPost
Turn a single URL, text snippet, or image into custom, platform-native social media posts without relying on third-party cloud tools. YominPost runs directly on your own machine, automatically picking the best post format for channels like X or LinkedIn, scoring its own copy, and running pre-publish risk checks to block prompt leaks or engagement bait before anything goes live. It hooks up to coding agents like Claude Code or Cursor through a local tool integration, letting AI assist with drafting while leaving you in full control of final approval.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ balanov/shim
The 7 Megabyte AI Terminal Tool
A single file under seven megabytes lets you control any artificial intelligence model, local or cloud, using plain language. This tool called Shim solves the headache of connecting different AI providers to your computer without installing heavy software. It acts like a universal adapter for your terminal. You can type simple requests or use specific commands to manage models and tools. The best part is that it runs entirely on your machine. It does not send your data to a central server. It keeps your work private and gives you full control over what the AI can access.
π @hackernewsgithubprojects
The 7 Megabyte AI Terminal Tool
A single file under seven megabytes lets you control any artificial intelligence model, local or cloud, using plain language. This tool called Shim solves the headache of connecting different AI providers to your computer without installing heavy software. It acts like a universal adapter for your terminal. You can type simple requests or use specific commands to manage models and tools. The best part is that it runs entirely on your machine. It does not send your data to a central server. It keeps your work private and gives you full control over what the AI can access.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ sqlpxc/airhand
Airhand: The Remote Desktop That Feels Local
Control any computer from your browser or desktop without installing heavy software. Airhand is a lightweight remote desktop tool that uses the QUIC protocol to get your screen to you faster than traditional options. The coolest part? It uses your computerβs graphics card to compress and send video, meaning almost zero lag even on shaky Wi-Fi. You can type, click, and hear system sounds in real-time, making it feel like you are sitting at the machine. Since it runs on Windows, Mac, and Linux, plus a web console, you can access your setup from anywhere with just a six-digit code.
π @hackernewsgithubprojects
Airhand: The Remote Desktop That Feels Local
Control any computer from your browser or desktop without installing heavy software. Airhand is a lightweight remote desktop tool that uses the QUIC protocol to get your screen to you faster than traditional options. The coolest part? It uses your computerβs graphics card to compress and send video, meaning almost zero lag even on shaky Wi-Fi. You can type, click, and hear system sounds in real-time, making it feel like you are sitting at the machine. Since it runs on Windows, Mac, and Linux, plus a web console, you can access your setup from anywhere with just a six-digit code.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ postgresgui/postgresgui
PostgresGUI: Your Native Mac Postgres Client
PostgresGUI is the native Mac client that keeps your database connections totally private. It solves the annoying problem of trusting web-based tools with your sensitive connection details by handling everything locally on your machine. Instead of routing your queries through some third-party server, it talks directly to your database. This means your connection info, actual queries, and the results you get back never leave your computer. It uses the macOS Keychain to securely store your passwords, which is a big win for security and peace of mind.
π @hackernewsgithubprojects
PostgresGUI: Your Native Mac Postgres Client
PostgresGUI is the native Mac client that keeps your database connections totally private. It solves the annoying problem of trusting web-based tools with your sensitive connection details by handling everything locally on your machine. Instead of routing your queries through some third-party server, it talks directly to your database. This means your connection info, actual queries, and the results you get back never leave your computer. It uses the macOS Keychain to securely store your passwords, which is a big win for security and peace of mind.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ justrach/nanolight
Nanolight: A 2 MB Launcher
Nanolight is the lightweight launcher that runs on pure Zig code, making it roughly two megabytes in size. It replaces heavy web-based apps with a fast panel that opens in a single keystroke. The standout feature is its ability to run Raycast extensions without needing the original Raycast app. It executes these scripts directly inside its own interface, giving you access to community tools while staying incredibly light. It also handles basic math and Google searches right there in the panel. This project proves that a launcher does not need to be bloated to be useful. It is a fast, open source alternative for anyone who wants speed over complexity.
π @hackernewsgithubprojects
Nanolight: A 2 MB Launcher
Nanolight is the lightweight launcher that runs on pure Zig code, making it roughly two megabytes in size. It replaces heavy web-based apps with a fast panel that opens in a single keystroke. The standout feature is its ability to run Raycast extensions without needing the original Raycast app. It executes these scripts directly inside its own interface, giving you access to community tools while staying incredibly light. It also handles basic math and Google searches right there in the panel. This project proves that a launcher does not need to be bloated to be useful. It is a fast, open source alternative for anyone who wants speed over complexity.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ zortos293/lichii
Lichii: The Early Switch 2 Emulator
Lichii is the experimental emulator bringing Nintendo Switch 2 games to your personal computer. This project solves the problem of playing next-generation hardware on older devices by recreating the console's internal processes in software. It is particularly interesting because it is in a very early stage, meaning developers are actively defining how it will handle graphics and performance right now. If you follow emulation closely, this is the project to watch as it matures from a rough prototype into a playable application. Keep an eye on it to see how the community builds the foundation for the next generation of home console emulation.
π @hackernewsgithubprojects
Lichii: The Early Switch 2 Emulator
Lichii is the experimental emulator bringing Nintendo Switch 2 games to your personal computer. This project solves the problem of playing next-generation hardware on older devices by recreating the console's internal processes in software. It is particularly interesting because it is in a very early stage, meaning developers are actively defining how it will handle graphics and performance right now. If you follow emulation closely, this is the project to watch as it matures from a rough prototype into a playable application. Keep an eye on it to see how the community builds the foundation for the next generation of home console emulation.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ mhfed/vn-law-check
VN Law Check: Audit Your Site
V N Law Check is the compliance agent that finally checks your website against actual Vietnamese law. It plugs directly into your coding tools like Cursor or Copilot to scan your site for real issues. The coolest part is that it does not just list rules, it looks at your actual code to see if you are breaking specific laws about data privacy or e-commerce. It then tells you exactly what to fix, like adding a proper consent form or updating an outdated privacy policy. It is not legal advice, but it is a huge head start for any developer building for Vietnam.
π @hackernewsgithubprojects
VN Law Check: Audit Your Site
V N Law Check is the compliance agent that finally checks your website against actual Vietnamese law. It plugs directly into your coding tools like Cursor or Copilot to scan your site for real issues. The coolest part is that it does not just list rules, it looks at your actual code to see if you are breaking specific laws about data privacy or e-commerce. It then tells you exactly what to fix, like adding a proper consent form or updating an outdated privacy policy. It is not legal advice, but it is a huge head start for any developer building for Vietnam.
π @hackernewsgithubprojects