This media is not supported in your browser
VIEW IN TELEGRAM
📦 nvlabs/cute
Understanding GPU Memory Layouts with CuTe
Design complex memory arrangements for high-performance computing without writing a single line of CUDA code. The cute repository provides a pure Python reference implementation of the hierarchical layout and tensor algebra used in Nvidia's high-performance computing libraries. It allows you to model how multi-dimensional data maps to physical memory addresses using simple, nested shapes and strides. You can experiment with complex matrix operations, visualize thread-to-value mappings, and design conflict-free memory patterns on your laptop with basic Python. It is an incredibly accessible way to master the advanced memory math that powers modern hardware acceleration.
📰 https://news.ycombinator.com/item?id=49103085
🆔 @hackernewsgithubprojects
Understanding GPU Memory Layouts with CuTe
Design complex memory arrangements for high-performance computing without writing a single line of CUDA code. The cute repository provides a pure Python reference implementation of the hierarchical layout and tensor algebra used in Nvidia's high-performance computing libraries. It allows you to model how multi-dimensional data maps to physical memory addresses using simple, nested shapes and strides. You can experiment with complex matrix operations, visualize thread-to-value mappings, and design conflict-free memory patterns on your laptop with basic Python. It is an incredibly accessible way to master the advanced memory math that powers modern hardware acceleration.
📰 https://news.ycombinator.com/item?id=49103085
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 kunal12203/graperoot
GrapeRoot AI Context Engine
Slash your AI coding bills and get accurate answers on the first try by feeding your developer tools exactly the context they need. GrapeRoot runs locally to build a smart semantic map of your entire codebase, tracing how your files, classes, and imports connect. Instead of making your AI assistant blindly search through thousands of files or wasting expensive tokens uploading redundant code, this tool pre-loads the precise code snippets required for your specific question. It even remembers what you edited across a session so your savings compound over time. Just run it alongside your favorite developer tool to get faster, cheaper, and smarter programming help.
🆔 @hackernewsgithubprojects
GrapeRoot AI Context Engine
Slash your AI coding bills and get accurate answers on the first try by feeding your developer tools exactly the context they need. GrapeRoot runs locally to build a smart semantic map of your entire codebase, tracing how your files, classes, and imports connect. Instead of making your AI assistant blindly search through thousands of files or wasting expensive tokens uploading redundant code, this tool pre-loads the precise code snippets required for your specific question. It even remembers what you edited across a session so your savings compound over time. Just run it alongside your favorite developer tool to get faster, cheaper, and smarter programming help.
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 alielsokary/caskhub
CaskHub
CaskHub is the native Mac app store that finally brings a beautiful interface to your command line setup. It acts as a full graphical store for Homebrew casks, letting you search, install, update, and uninstall thousands of Mac apps without ever touching the terminal. The coolest part is its safe adoption feature, which lets you take apps you already downloaded from websites or the App Store and bring them under command line management with a single click. It is completely free, open source, and preserves full terminal compatibility, making app management incredibly satisfying.
🆔 @hackernewsgithubprojects
CaskHub
CaskHub is the native Mac app store that finally brings a beautiful interface to your command line setup. It acts as a full graphical store for Homebrew casks, letting you search, install, update, and uninstall thousands of Mac apps without ever touching the terminal. The coolest part is its safe adoption feature, which lets you take apps you already downloaded from websites or the App Store and bring them under command line management with a single click. It is completely free, open source, and preserves full terminal compatibility, making app management incredibly satisfying.
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 tencent/angelspec
AngelSpec Speculative Decoding
AngelSpec is the training framework that finally unifies speculative decoding models under a single, highly efficient pipeline. Speculative decoding speeds up large language models by letting a smaller, faster draft model guess the next few tokens, which the main model then quickly verifies. Instead of forcing you to build custom setups for every new draft method, this tool brings six different major architectures together, letting you switch between them with a simple configuration tweak. It connects your training and generation processes across separate graphics chips using a clever shared memory store, so they can scale completely independently. If you want to make your language models run significantly faster in production...
🆔 @hackernewsgithubprojects
AngelSpec Speculative Decoding
AngelSpec is the training framework that finally unifies speculative decoding models under a single, highly efficient pipeline. Speculative decoding speeds up large language models by letting a smaller, faster draft model guess the next few tokens, which the main model then quickly verifies. Instead of forcing you to build custom setups for every new draft method, this tool brings six different major architectures together, letting you switch between them with a simple configuration tweak. It connects your training and generation processes across separate graphics chips using a clever shared memory store, so they can scale completely independently. If you want to make your language models run significantly faster in production...
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 achrefelouafi/grasssystemthreejs
Grass System Three JS
Build highly detailed and responsive environment scenes right in the browser. Grass System Three JS lets you sculpt real-time, procedural terrain directly on the GPU using Three JS. Instead of loading heavy static meshes, this sandbox dynamically generates mounds, mossy layers, and dry cracks using advanced math shaders. Every single blade of grass is instanced and animated in a single draw call, reacting to custom wind gusts and matching the contours of your terrain. You can even drop in 3D models and watch virtual moss grow over them. It is an incredibly clever way to build lightweight, interactive 3D landscapes that run fast on almost any screen.
🆔 @hackernewsgithubprojects
Grass System Three JS
Build highly detailed and responsive environment scenes right in the browser. Grass System Three JS lets you sculpt real-time, procedural terrain directly on the GPU using Three JS. Instead of loading heavy static meshes, this sandbox dynamically generates mounds, mossy layers, and dry cracks using advanced math shaders. Every single blade of grass is instanced and animated in a single draw call, reacting to custom wind gusts and matching the contours of your terrain. You can even drop in 3D models and watch virtual moss grow over them. It is an incredibly clever way to build lightweight, interactive 3D landscapes that run fast on almost any screen.
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 grinev/opencode-telegram-bot
OpenCode Telegram Bot
You can now run and monitor local artificial intelligence coding tasks directly from your phone without exposing any APIs or opening public network ports. The open source project opencode-telegram-bot establishes a secure connection between your mobile device and your local computer through the Telegram API. It acts as an interactive client for the OpenCode engine, letting you send prompts, switch models, and approve agent permissions via simple chat buttons. Since it is entirely self-hosted and relies on a strict user whitelist, your development environment remains private and safe. It is the perfect bridge for developers who want to step away from their desks while keeping their local code generation running.
🆔 @hackernewsgithubprojects
OpenCode Telegram Bot
You can now run and monitor local artificial intelligence coding tasks directly from your phone without exposing any APIs or opening public network ports. The open source project opencode-telegram-bot establishes a secure connection between your mobile device and your local computer through the Telegram API. It acts as an interactive client for the OpenCode engine, letting you send prompts, switch models, and approve agent permissions via simple chat buttons. Since it is entirely self-hosted and relies on a strict user whitelist, your development environment remains private and safe. It is the perfect bridge for developers who want to step away from their desks while keeping their local code generation running.
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 nianzhibai/91
91 Video Station
The media server that automatically connects and deduplicates your cloud storage drives into a single unified streaming interface. Nine One mounts popular services like OneDrive, WebDAV, and local storage, cataloging your entire media collection into a sleek React frontend. It handles the heavy lifting behind the scenes, using ffmpeg to extract thumbnails and generate multi-segment preview teasers for every clip, and it even includes an automated duplicate detector that runs content-level frame matching to keep your storage clean. It is the perfect self-hosted solution for spinning up a private Netflix-like dashboard directly from your cloud files.
🆔 @hackernewsgithubprojects
91 Video Station
The media server that automatically connects and deduplicates your cloud storage drives into a single unified streaming interface. Nine One mounts popular services like OneDrive, WebDAV, and local storage, cataloging your entire media collection into a sleek React frontend. It handles the heavy lifting behind the scenes, using ffmpeg to extract thumbnails and generate multi-segment preview teasers for every clip, and it even includes an automated duplicate detector that runs content-level frame matching to keep your storage clean. It is the perfect self-hosted solution for spinning up a private Netflix-like dashboard directly from your cloud files.
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 adithyan-ak/agenthound
Agent Hound
Agent Hound is the security tool that maps out every hidden vulnerability inside your AI agent infrastructure. Think of it as bloodhound for artificial intelligence, scanning everything from model gateways to local prompt configurations to find exactly where your system is exposed. Instead of guessing how a rogue agent could pivot through your network, it merges your entire setup into a single interactive graph that shows you real attack paths, credential leaks, and poisoning risks. It lets you visualize exactly how a compromised tool could lead to a massive data leak before it actually happens. Check out Agent Hound on GitHub to map and secure your AI playground today.
📰 https://news.ycombinator.com/item?id=49091338
🆔 @hackernewsgithubprojects
Agent Hound
Agent Hound is the security tool that maps out every hidden vulnerability inside your AI agent infrastructure. Think of it as bloodhound for artificial intelligence, scanning everything from model gateways to local prompt configurations to find exactly where your system is exposed. Instead of guessing how a rogue agent could pivot through your network, it merges your entire setup into a single interactive graph that shows you real attack paths, credential leaks, and poisoning risks. It lets you visualize exactly how a compromised tool could lead to a massive data leak before it actually happens. Check out Agent Hound on GitHub to map and secure your AI playground today.
📰 https://news.ycombinator.com/item?id=49091338
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 cvar-vision-dl/remind-reid-tracker
remind-reid-tracker
This computer vision system can accurately re-identify and track indoor objects even after they completely disappear from view, experience massive lighting changes, or are seen from entirely new angles. Named remind-reid-tracker, this online tracker solves a major headache for mobile robots by building persistent, adaptive memory models of every object using frozen deep learning features. Instead of relying on predictable motion paths, it analyzes semantic parts and relational context, like how objects are arranged next to each other. When things get too visually confusing, it intelligently pauses and flags those objects as ambiguous instead of making a wrong guess.
🆔 @hackernewsgithubprojects
remind-reid-tracker
This computer vision system can accurately re-identify and track indoor objects even after they completely disappear from view, experience massive lighting changes, or are seen from entirely new angles. Named remind-reid-tracker, this online tracker solves a major headache for mobile robots by building persistent, adaptive memory models of every object using frozen deep learning features. Instead of relying on predictable motion paths, it analyzes semantic parts and relational context, like how objects are arranged next to each other. When things get too visually confusing, it intelligently pauses and flags those objects as ambiguous instead of making a wrong guess.
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 orchidfiles/ungate
Ungate
Route your Cursor chat requests directly through your web subscriptions to stop paying for expensive API tokens. By setting up a secure local proxy and a Cloudflare tunnel, ungate intercepts your editor requests and translates them so they run seamlessly under your existing Claude and ChatGPT account plans. It handles everything behind the scenes, from translating complex streaming tool calls to fixing that annoying Cursor bug where your API key settings randomly turn off. You get the same smart completions and image support in your editor without running up a massive developer bill. Just install the extension, link your accounts, and write code on your subscription's dime.
📰 https://news.ycombinator.com/item?id=48135776
🆔 @hackernewsgithubprojects
Ungate
Route your Cursor chat requests directly through your web subscriptions to stop paying for expensive API tokens. By setting up a secure local proxy and a Cloudflare tunnel, ungate intercepts your editor requests and translates them so they run seamlessly under your existing Claude and ChatGPT account plans. It handles everything behind the scenes, from translating complex streaming tool calls to fixing that annoying Cursor bug where your API key settings randomly turn off. You get the same smart completions and image support in your editor without running up a massive developer bill. Just install the extension, link your accounts, and write code on your subscription's dime.
📰 https://news.ycombinator.com/item?id=48135776
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 superxslam/supermap
SuperMap
Imagine a robot that does not just map a room, but actually remembers how it changes over time. Carnegie Mellon researchers built supermap, a living spatial memory that acts like a four-dimensional brain for AI. Instead of static 3D models, it builds a queryable timeline of every single object, keeping track of its location even when it gets blocked from view. A robot can use this memory to look at a room, understand where a coffee cup went, and reason through complex tasks. It runs fully live on robot hardware, making real-world navigation genuinely smart.
🆔 @hackernewsgithubprojects
SuperMap
Imagine a robot that does not just map a room, but actually remembers how it changes over time. Carnegie Mellon researchers built supermap, a living spatial memory that acts like a four-dimensional brain for AI. Instead of static 3D models, it builds a queryable timeline of every single object, keeping track of its location even when it gets blocked from view. A robot can use this memory to look at a room, understand where a coffee cup went, and reason through complex tasks. It runs fully live on robot hardware, making real-world navigation genuinely smart.
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 smixs/pohuy
pohuy
Imagine an artificial intelligence assistant that actually talks like a real senior software engineer who has spent twenty grueling years fixing production servers. This brilliant open-source project called pohuy introduces an idiomatic Russian profanity mode for artificial intelligence agents. It replaces polite corporate filler with incredibly accurate, context-aware engineering slang, categorizing project emergencies from simple minor bugs all the way to complete data disasters. It keeps raw code and pull requests clean, but lets the chat run wild with authentic developer energy. It is the perfect way to make debugging feel like you are talking to a real human friend.
🆔 @hackernewsgithubprojects
pohuy
Imagine an artificial intelligence assistant that actually talks like a real senior software engineer who has spent twenty grueling years fixing production servers. This brilliant open-source project called pohuy introduces an idiomatic Russian profanity mode for artificial intelligence agents. It replaces polite corporate filler with incredibly accurate, context-aware engineering slang, categorizing project emergencies from simple minor bugs all the way to complete data disasters. It keeps raw code and pull requests clean, but lets the chat run wild with authentic developer energy. It is the perfect way to make debugging feel like you are talking to a real human friend.
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 vladuzh/qwen-scribe
Qwen Scribe
Your Mac can now transcribe audio and video locally without sending a single byte of data to the cloud. A clever developer built qwen scribe, a lightweight desktop app designed for Apple Silicon that uses a local open-source speech model to handle completely private transcription and system-wide dictation on your GPU. By simply holding down the right Command key in any application, you can speak, release the key, and watch your words instantly paste right at your cursor. It is the perfect private alternative to cloud dictation, keeping your voice recordings and transcripts safely locked on your own machine.
📰 https://news.ycombinator.com/item?id=49098260
🆔 @hackernewsgithubprojects
Qwen Scribe
Your Mac can now transcribe audio and video locally without sending a single byte of data to the cloud. A clever developer built qwen scribe, a lightweight desktop app designed for Apple Silicon that uses a local open-source speech model to handle completely private transcription and system-wide dictation on your GPU. By simply holding down the right Command key in any application, you can speak, release the key, and watch your words instantly paste right at your cursor. It is the perfect private alternative to cloud dictation, keeping your voice recordings and transcripts safely locked on your own machine.
📰 https://news.ycombinator.com/item?id=49098260
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 mpociot/codex-micro-stream-deck-emulator
Codex Micro Stream Deck Emulator
You can turn your ordinary Elgato Stream Deck into a custom physical controller for the ChatGPT desktop app. With the clever codex-micro-stream-deck-emulator, the ChatGPT app gets tricked into thinking your Stream Deck is a genuine Codex Micro macro pad. When you connect it, your device keys instantly reflect your active tasks with color-coded status updates, while the built-in action keys and dials control reasoning depth and push-to-talk mic features. It achieves this magic on macOS by running a background bridge and injecting a tiny, non-invasive runtime shim into the ChatGPT app. It is the ultimate productivity hack for power users who want tactile, real-time physical control over their AI workflows.
🆔 @hackernewsgithubprojects
Codex Micro Stream Deck Emulator
You can turn your ordinary Elgato Stream Deck into a custom physical controller for the ChatGPT desktop app. With the clever codex-micro-stream-deck-emulator, the ChatGPT app gets tricked into thinking your Stream Deck is a genuine Codex Micro macro pad. When you connect it, your device keys instantly reflect your active tasks with color-coded status updates, while the built-in action keys and dials control reasoning depth and push-to-talk mic features. It achieves this magic on macOS by running a background bridge and injecting a tiny, non-invasive runtime shim into the ChatGPT app. It is the ultimate productivity hack for power users who want tactile, real-time physical control over their AI workflows.
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 digimata/quill
Quill
Record and transcribe your meetings completely offline using a tiny, ultra-minimalist menu bar app for Mac. Quill runs entirely on your local machine, capturing your microphone and your computer's system audio as two separate tracks. When you stop recording, it automatically runs an on-device speech model to generate a speaker-tagged transcript, matching your voice against the other side of the call with zero data ever leaving your computer. By separating the audio streams, it achieves incredibly accurate transcriptions without needing complex speaker-identification models. It is the perfect, private way to document your calls and keep your data safe.
🆔 @hackernewsgithubprojects
Quill
Record and transcribe your meetings completely offline using a tiny, ultra-minimalist menu bar app for Mac. Quill runs entirely on your local machine, capturing your microphone and your computer's system audio as two separate tracks. When you stop recording, it automatically runs an on-device speech model to generate a speaker-tagged transcript, matching your voice against the other side of the call with zero data ever leaving your computer. By separating the audio streams, it achieves incredibly accurate transcriptions without needing complex speaker-identification models. It is the perfect, private way to document your calls and keep your data safe.
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 yoanwai/agent-manager
Agent Manager
Agent Manager is the terminal user interface that finally lets you run and monitor all your artificial intelligence coding agents side by side. It coordinates tools like Claude Code, Codex, and Grok Build in independent terminal multiplexer sessions, ensuring they keep running in the background even if you close the dashboard. Instead of digging through endless terminal tabs to see if an assistant is stuck or finished, you get a clean, grouped project tree showing live statuses and file previews. Best of all, you can review exact code diffs and send feedback directly to an agent with a single keystroke.
📰 https://news.ycombinator.com/item?id=49107749
🆔 @hackernewsgithubprojects
Agent Manager
Agent Manager is the terminal user interface that finally lets you run and monitor all your artificial intelligence coding agents side by side. It coordinates tools like Claude Code, Codex, and Grok Build in independent terminal multiplexer sessions, ensuring they keep running in the background even if you close the dashboard. Instead of digging through endless terminal tabs to see if an assistant is stuck or finished, you get a clean, grouped project tree showing live statuses and file previews. Best of all, you can review exact code diffs and send feedback directly to an agent with a single keystroke.
📰 https://news.ycombinator.com/item?id=49107749
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 amd-agi/instella-moe
Instella MoE
This open source project trained a state-of-the-art, sixteen-billion-parameter language model completely from scratch on advanced silicon. The repository contains the complete training pipeline for instella-moe, a mixture-of-experts model that only activates a fraction of its parameters at any given moment to keep generation fast and efficient. It guides you step-by-step through every phase of development, including pre-training, long-context extension, and direct preference optimization. Developers can use this codebase to easily reproduce or build upon high-performance language systems on modern hardware. Check out the repository to run the code, download the pre-trained weights, and explore the advanced architecture today.
🆔 @hackernewsgithubprojects
Instella MoE
This open source project trained a state-of-the-art, sixteen-billion-parameter language model completely from scratch on advanced silicon. The repository contains the complete training pipeline for instella-moe, a mixture-of-experts model that only activates a fraction of its parameters at any given moment to keep generation fast and efficient. It guides you step-by-step through every phase of development, including pre-training, long-context extension, and direct preference optimization. Developers can use this codebase to easily reproduce or build upon high-performance language systems on modern hardware. Check out the repository to run the code, download the pre-trained weights, and explore the advanced architecture today.
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 artexis10/endstate-gui
endstate-gui
Setting up a brand new computer usually means wasting hours downloading installers and adjusting buried settings. With endstate-gui, you can package your entire digital setup into a single portable file and restore everything in moments. It captures your installed applications and configurations, then rolls them out onto a clean machine with just a few clicks. The desktop interface provides clear visual summaries of your system state, lets you choose exactly which settings to migrate, and tracks the restoration process in real time. It is a simple way to make upgrading your hardware or resetting your workspace completely stress-free.
🆔 @hackernewsgithubprojects
endstate-gui
Setting up a brand new computer usually means wasting hours downloading installers and adjusting buried settings. With endstate-gui, you can package your entire digital setup into a single portable file and restore everything in moments. It captures your installed applications and configurations, then rolls them out onto a clean machine with just a few clicks. The desktop interface provides clear visual summaries of your system state, lets you choose exactly which settings to migrate, and tracks the restoration process in real time. It is a simple way to make upgrading your hardware or resetting your workspace completely stress-free.
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 gimanh/taskview-community
TaskView Community
Manage your software projects on your own infrastructure without the bloat of massive enterprise platforms. This source-available project management tool puts you back in complete control of your data, providing a clean alternative to subscription-based services. You can easily organize tasks with visual boards, track deep dependencies, and monitor sprints, all while keeping your team's sensitive information self-hosted. The most exciting part is the built-in model context protocol server, which lets you connect AI assistants to securely update statuses and automate workflows. It is the perfect way to run private, intelligent project management on your own terms.
📰 https://news.ycombinator.com/item?id=49115078
🆔 @hackernewsgithubprojects
TaskView Community
Manage your software projects on your own infrastructure without the bloat of massive enterprise platforms. This source-available project management tool puts you back in complete control of your data, providing a clean alternative to subscription-based services. You can easily organize tasks with visual boards, track deep dependencies, and monitor sprints, all while keeping your team's sensitive information self-hosted. The most exciting part is the built-in model context protocol server, which lets you connect AI assistants to securely update statuses and automate workflows. It is the perfect way to run private, intelligent project management on your own terms.
📰 https://news.ycombinator.com/item?id=49115078
🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 perryts/perry
Perry
Your TypeScript code can now run as fast as Rust without needing a JavaScript engine at all. A clever new native compiler called Perry compiles TypeScript directly into standalone, machine-code executables. It completely bypasses Node, Bun, and Electron, stripping away the bulky runtime entirely. By utilizing SWC and LLVM, Perry produces tiny, self-contained binaries that start instantly, with a basic hello world landing at just three hundred thirty kilobytes. It even links to real native platform widgets for building lightweight user interfaces. If you want high-performance, single-digit megabyte applications using the language you already know, it is time to try compiling your TypeScript.
📰 https://news.ycombinator.com/item?id=49115217
🆔 @hackernewsgithubprojects
Perry
Your TypeScript code can now run as fast as Rust without needing a JavaScript engine at all. A clever new native compiler called Perry compiles TypeScript directly into standalone, machine-code executables. It completely bypasses Node, Bun, and Electron, stripping away the bulky runtime entirely. By utilizing SWC and LLVM, Perry produces tiny, self-contained binaries that start instantly, with a basic hello world landing at just three hundred thirty kilobytes. It even links to real native platform widgets for building lightweight user interfaces. If you want high-performance, single-digit megabyte applications using the language you already know, it is time to try compiling your TypeScript.
📰 https://news.ycombinator.com/item?id=49115217
🆔 @hackernewsgithubprojects