56 subscribers
7.28K videos
7.88K links
Download Telegram
This media is not supported in your browser
VIEW IN TELEGRAM
📦 baidu-baige/loongflow

LoongFlow AI Agent Framework

Build expert-grade AI systems that actually learn from their own mistakes instead of just repeating them. LoongFlow organizes artificial intelligence into a structured planning, execution, and summary loop, meaning the agent constantly reflects on what went wrong and how to fix it. It even uses specialized memory to store and reuse these hard-won insights across complex coding, machine learning, and math tasks. Think of it as giving your local AI model the rigorous, reflective mindset of a human scientist. You can use it to build incredibly stable, self-improving agents that tackle your toughest problems with genuine, evolving logic.

📰 https://news.ycombinator.com/item?id=49049823

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 chinmay-sawant/gopdfsuit

gopdfsuit

Full compliance archival PDF generation delivers an astonishing sixty-six hundred operations per second on a single machine, while turning off compliance reaches an unbelievable thirty-seven thousand operations per second. This incredible speed comes from gopdfsuit, an open-source engine that works as a fast Go library, a REST API, or Python bindings. It generates fully tagged, digitally signed, and accessible documents directly from JSON templates, completely in-memory with zero external runtime dependencies. It is the perfect modern toolkit for any developer needing to quickly generate, merge, secure, or redact highly compliant documents at scale. Check out the gopdfsuit repository to speed up your document workflows today.

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 jbwinters/jacquard-lang

Jacquard Lang

Run and verify untrusted code written by machine learning models without risking your system. This project introduces a small programming language called jacquard-lang that is built specifically for a world where AI writes the code and humans review it. Instead of guessing what a model-generated script might do, this language explicitly lists every outside-world effect, like network or file access, directly in the function signature. If an effect isn't declared, the code won't compile, and if it isn't granted at runtime, it won't run. It is the ultimate safety net for reviewing AI code with absolute peace of mind.

📰 https://news.ycombinator.com/item?id=48894630

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 naver-ai/opd2

On-Policy Delta Distillation

Teach smaller artificial intelligence models how to reason by capturing only the specific changes that occur during a larger model's training. This approach uses a technique called on-policy delta distillation, which isolates the exact reasoning shift between a smart teacher model and its basic version. Instead of copying everything, the system focuses solely on this difference, or delta signal. Tests show this method consistently boosts performance in math, science, and coding tasks across various model sizes. It offers a much smarter, more direct way to pass down complex thinking skills to lightweight models without the usual extra bloat.

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 34306/usbliter8-fun

usbliter8-fun

usbliter8-fun is the experimental developer tool that lets you jailbreak and run custom firmware on an iPhone 11 Pro running iOS 27 beta. By hooking up a cheap Raspberry Pi Pico 2 board to act as a hardware exploit injector, you can bypass USB restrictions, patch the iOS kernel, and run unsigned code. The coolest part is that it lets you share your Mac's internet connection over USB and spin up a local VNC server, giving you full remote control of your phone's screen directly from your desktop. It is a brilliant, hands-on playground for developers who love pushing hardware to its absolute limits.

📰 https://news.ycombinator.com/item?id=48990541

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 tobi/try

Try CLI

Your local storage is likely cluttered with dozens of abandoned test directories, but a minimalist command line tool named try completely solves this file system chaos. Built as a single file Ruby script with zero dependencies, this utility acts as a smart, dedicated home for all your chaotic coding experiments. It automatically prefixes new directories with the current date, lets you clone git repositories straight into fresh folders, and uses a clever fuzzy search algorithm to instantly jump to your past work. The tool even tracks when you last opened a folder so your active projects naturally bubble up to the top. Install try today to keep your ideas organized...

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 l3lackcurtains/trading-ops

Trading Ops

Trading-ops is the terminal-based trading workspace that transforms Claude Code into a fully systematic market scanner. By running simple slash commands, you can instantly analyze stocks, crypto, foreign exchange, indices, or commodities, pulling back structured markdown verdicts, neat text-based price ladders, and clear trade tables. Instead of relying on vague vibes, the system runs local python scripts to gather hard data like volume profiles and moving averages without needing expensive paid subscriptions. It saves every analysis as a local file, creating a reliable audit trail of your trading decisions that you can easily track and automate over time.

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 franckyb/comfyui-prompt-manager

ComfyUI Prompt Manager

You can now run a local artificial intelligence model right inside ComfyUI to write and polish your prompts for you. With comfyui-prompt-manager, a local llama dot c p p or Ollama instance hooks directly into your workspace, automatically downloading optimized models to enhance your text or even analyze images to write descriptive prompts. Beyond AI generation, it acts as a central hub where you can save complex prompts, manage dual LoRA stacks, and extract complete generation settings straight from existing images or videos. It is the ultimate tool for organizing your creative chaos and building repeatable image recipes without messy node wiring.

🆔 @hackernewsgithubprojects
Media is too big
VIEW IN TELEGRAM
📦 pana1v/polka

Polka Multi LiDAR Fusion for ROS 2

Combine multiple lidar and laser scanner feeds into a single clean data stream with Polka. If you run multiple sensors on a robot, you usually have to daisy chain separate software nodes to filter, transform, and merge the data, which adds lag and wastes processing power. Polka replaces that entire messy chain with one highly efficient node. It merges different sensor formats, applies customizable filters, and can even use graphics card acceleration to speed up the process. A smart algorithm also uses motion data to instantly correct any sensor blur caused by the robot moving. The result is a single, clean, lightweight output topic that makes life much easier for...

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 hahhforest/pi-textbook

Build Your Own Pi

You can now build a fully functional coding assistant from scratch by following fifteen real checkpoints in a shared git history. The pi-textbook project is an interactive, executable engineering textbook written in Chinese that guides developers through building a Pi-style agent from first principles. Instead of showing you theoretical pseudocode, this course connects every single chapter to an actual runnable code commit, focused test suites, and controlled failure experiments. You will build the system step-by-step, starting with a single offline trace and working up to streaming model events, custom coding tools, session trees, and runtime evaluation. It is the ultimate hands-on masterclass for understanding how modern AI agents actually run.

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 dotgui/core

Core UI Format

User interfaces are finally getting their own universal image format. Think about how portable vector graphics are because of scalable vector graphics, but those were made for illustrations, not modern app screens. A clever new open-source standard called core introduces the dot-gui format, a human-readable, text-based file structure built specifically for layouts, components, and design tokens. Instead of locking designs inside proprietary platforms or dealing with complex web-bundler friction, dot-gui organizes your interfaces as clean packages that any code renderer, AI model, or design tool can read and write. It is the missing link that lets your favorite design tools finally talk to each other directly.

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 l0ng-ai/tty7

tty7 Terminal

Keep your terminal shells running even after a system reboot without ever needing to configure tmux. The project is called tty7, a lightning-fast terminal workbench built in Rust that completely reimagines your command-line workflow. By using a persistent background daemon, it keeps all your active sessions, tabs, and splits alive through restarts, so you can pick up right where you left off. It combines Alacritty's terminal core with GPU rendering to double the speed of traditional modern terminals, while embedding smart history suggestions, syntax highlighting, and native SSH profiles directly into your workspace. It is the ultimate command center that makes your terminal feel like a modern editor.

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 regolo-ai/brick-sr1

Brick SR1

You can now run every major language model behind a single smart endpoint and let an intelligent routing gateway instantly pick the cheapest backend that can actually handle your prompt. Instead of wasting time and money sending simple questions to giant models or cascading through multiple tries, this project analyzes the complexity and required skills of your query to make a single, precise dispatch. It works seamlessly as an open source proxy for coding assistants like Claude Code and Codex, letting you dynamically switch between light and heavy models without losing context or wasting cache memory. Try setting up the router to slash your development API bills while keeping the...

📰 https://news.ycombinator.com/item?id=48865822

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 bitmiracle-ai/dormice

Dormice

Most cloud platforms charge you by the second for running AI agent sandboxes, forcing you to delete them. Dormice completely flips this model by letting you run permanent sandboxes on your own machine that cost absolutely nothing when idle. When an agent is inactive, the system automatically cools the sandbox down, shrinking its memory from a whole gigabyte to just five megabytes, and then wakes it back up in fifty milliseconds the moment you need it. It is compatible with standard packages, runs on a single server with zero complex setup, and even archives old data to cloud storage automatically. Now you can let your agents run forever without worrying...

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 vegetebird/vlm-ie3d

vlm-ie3d

The visual language model framework vlm-ie3d is the open-source software that bridges the gap between high-level scene understanding and fine-grained spatial mapping. Traditional visual language models struggle to grasp exact spatial details because they rely entirely on rough cognitive estimates. This framework solves the problem by blending coarse implicit geometries with explicit, reconstructed three-dimensional structural details extracted directly from standard two-dimensional video. By combining these two complementary layers of spatial information through a specialized adapter, it gives artificial intelligence a much sharper, more precise grasp of depth and physical surroundings. It is a massive step forward for anyone training models to navigate and understand physical spaces.

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 larashero3-dotcom/lieflat-charts

Lieflat Charts

You can now turn raw data into editorial-grade interactive charts instantly using a single, beautiful HTML file. The lieflat-charts project is a template-driven monochrome data visualization tool built specifically for artificial intelligence agents. Instead of generic, colorful dashboards, it crafts sophisticated, publication-ready layouts using a strict palette of rich charcoal and warm paper-gray. It guides agents through a rigorous decision tree to match your data shape with the perfect storytelling format, from detailed hairline plots to dense unit grids. Best of all, everything renders in a standalone HTML file with clean inline styles, requiring absolutely no build steps or complex dependencies.

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 huster-hq/m2c

M2C

M2C is the fine tuning strategy that finally makes image annotation incredibly fast and efficient with minimal data. If you have ever tried to train a computer vision model, you know that labeling images is a total chore, especially when you only have one or two examples to work with. This project solves that by introducing an efficient search method that works with Meta's Segment Anything Model 3, allowing you to run highly accurate few shot image segmentation. This means you can automatically generate perfect masks for complex medical images or custom datasets with almost zero manual effort, saving hours of tedious clicking. It is like having an AI labeling...

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 blackmixture/mix-studio

Mix Studio: The Cleanest AI Workspace

Run highly tuned image and video workflows flawlessly from your desktop or your phone using Mix Studio. This local web interface completely transforms how you use ComfyUI by handling all the complex node graph construction behind the scenes. Instead of getting lost in a spaghetti mess of nodes and wires, you work through clean, touch-friendly controls while the application automatically builds and submits the proper API graphs for you. It lets you run advanced image generation, regional prompting, multi-reference editing, and video motion transfer locally. Best of all, because it runs on your local machine, your prompts, inputs, and generated media stay private.

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 h3pdesign/neon-vision-editor

Neon Vision Editor

While typical code editors crush your laptop under gigabytes of memory-hogging bloat, neon-vision-editor is a beautifully lightweight, native text editor built strictly for Apple devices. It brings lightning-fast startup and smooth tabbed editing to macOS, iPhone, and iPad without tracking your data or packing a massive development environment you do not need. The coolest part is how it handles giant documents. If you open a massive config file, it automatically switches into a tailored large file mode that disables heavy syntax analysis to keep your scrolling and typing perfectly fluid. It is the ideal, no-nonsense tool when you just want to write some markdown, tweak a script, or take quick...

🆔 @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
📦 perplexityai/api-platform-developers

API Platform Developers

The Perplexity API platform developers repository publishes portable agent skills that let you migrate codebases from the old Sonar chat completions to the new Agent API with absolute precision. This project is incredibly useful because the new Agent API is in strict-mode, meaning it will throw a four-hundred error if your request contains even a single leftover parameter from the legacy API. It packs specific skills for tools like Claude Code and Codex CLI, giving your development agents the exact recipes and verification scripts they need to swap endpoints, handle streaming events, and map web search filters without breaking your production apps. It is the ultimate automation shortcut for keeping...

🆔 @hackernewsgithubprojects