56 subscribers
7.26K videos
7.87K links
Download Telegram
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ yassa9/dvlt.cu

3D Reconstruction on Your GPU Without the Bloat

Ever wanted to turn a handful of images into a 3D point cloud without installing massive, bloated frameworks? Meet this project, a lightning-fast, zero-dependency CUDA implementation that brings NVIDIA's looping transformer technology straight to your consumer GPU. It ditches Python and heavy libraries entirely, offering a single five-megabyte binary that reconstructs static scenes in one forward pass. By utilizing fused kernels and efficient memory management, it handles complex 3D reconstruction tasks with ease. Whether you are tired of lengthy training times or just love clean, high-performance code, this project is a perfect look at how lean your software can actually be.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ paperguru-ai/paperguru-benchmark

Is Your AI Forgetful? Meet PaperGuru

Are your long-horizon AI agents struggling to remember key details? Most systems rely on basic retrieval, but PaperGuru introduces a game-changing solution called Lifecycle-Aware Memory. This architecture moves beyond standard methods by tracking versioned content and structural relationships, ensuring AI agents stay accurate even as information evolves. By utilizing a capital chunk memory mechanism, it effectively routes information through a temporal graph to deliver precise, provenance-grounded evidence. With state-of-the-art results on major benchmarks like PaperBench and SurveyBench, this project provides the essential memory primitive modern agents are missing. Level up your AI memory and keep your agents consistent today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ cuevhv/mamma

Capture Human Motion with MAMMA

Ever wondered how to capture complex human movement without wearing bulky suits? Meet MAMMA, a powerful new tool designed for markerless motion acquisition that turns multi-camera footage into precise 3D human models. By combining advanced segmentation with custom landmark detection and SMPL-X optimization, this pipeline accurately tracks multiple people in diverse settings. Whether you are a researcher or a creator, you can use the command line or the intuitive browser interface to process your own footage with ease. It is a fantastic step forward for anyone looking to bring realistic motion into their digital projects today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ mathiasconradt/tektite

Tektite: A Minimalistic Note-Taking App for macOS and Linux

Are you tired of bloated note-taking apps that track your every move? Meet Tektite, a minimalistic, open-source knowledge base designed specifically for macOS and Linux. It puts you back in control by storing your notes as simple local files on your machine. With features like live Markdown preview, a visual graph to map your thoughts, and built-in support for Git synchronization, it is the perfect tool for keeping your data private and portable. No accounts, no telemetry, and no unnecessary pluginsβ€”just pure, focused writing. Download Tektite today and experience a clean, distraction-free way to organize your digital world.

πŸ“° https://news.ycombinator.com/item?id=48278518

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ braedonsaunders/codeflow

Visualize Your Codebase Instantly with CodeFlow

Ever felt lost navigating a massive new codebase? CodeFlow solves this by transforming any GitHub repository or local project into an interactive architecture map in seconds. This privacy-first tool runs entirely in your browser, meaning your code never leaves your machine. You can explore dependency graphs, identify security vulnerabilities, and perform blast radius analysis to see exactly how changes might impact your files. It even detects design patterns and grades your project health automatically. Whether you are reviewing PRs or mapping your own project, CodeFlow makes understanding complex architecture effortless. Stop guessing and start seeing your code clearly today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ exon-research/genomi

Turn Your AI Into a DNA Expert

Ever wondered what your DNA actually says about your health, nutrition, or medication response? Decoding your genome is overwhelming, with billions of base pairs to analyze. That is where Genomi changes the game. It is an open-source AI agent runtime that transforms your AI into a private, personal DNA expert. By creating a local Active Genome Index, it allows your agent to query genetic evidence while keeping your sensitive data securely on your own machine. It works with most MCP-capable agents, providing evidence-backed answers to your health questions. Unlock the secrets hidden in your biology today with this powerful tool.

πŸ“° https://news.ycombinator.com/item?id=48384270

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ facebookresearch/projectaria_gen2_depth_from_stereo

Turn Aria Gen2 Video Into 3D Depth Maps

Ever wonder how to extract precise 3D depth from your Aria Gen2 recordings? This project provides a powerful toolkit designed to transform your raw stereo camera data into detailed metric depth maps. By using stereo rectification and the Foundation Stereo model, it automates the entire pipeline from loading files to generating clean output. You get flexible command-line utilities and a step-by-step notebook, allowing you to choose between PyTorch or faster TensorRT backends for your inference needs. It is an essential resource for anyone looking to bridge the gap between simple video footage and high-quality spatial data for downstream projects.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ microsoft/pg_durable

Stop Stitching Workflow Tools With Postgres Durable Execution

Are you tired of managing complex background workers, cron jobs, and queues just to handle reliable data tasks? Meet pg_durable, a PostgreSQL extension that brings durable execution directly inside your database. It allows you to define long-running workflows as a graph of SQL steps that checkpoint automatically. If your database crashes or a step fails, execution resumes exactly where it left off instead of restarting from scratch. By keeping your orchestration logic and state within Postgres, you eliminate external infrastructure and simplify your architecture. It is a powerful way to make complex pipelines resilient, auditable, and much easier to manage.

πŸ“° https://news.ycombinator.com/item?id=48414367

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ hzxie/dynamicvla

Mastering Robot Manipulation with DynamicVLA

Ever wonder how robots learn to handle moving objects with precision? Meet DynamicVLA, an advanced vision-language-action model designed specifically for complex dynamic object manipulation. This repository provides the official implementation, giving you the tools to train and test policies that adapt to changing environments. By utilizing the included DOM dataset and integrating with Isaac Lab for realistic simulation, developers can generate synthetic data, evaluate performance, and refine robotic control. It is a powerful framework for researchers looking to bridge the gap between visual input and physical action. Dive into the code and start building the future of robotics today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ travisvn/awesome-claude-skills

Supercharge Your Claude Workflows with These Skills

Do you want to turn Claude into a powerhouse for your specific projects? This repository is a curated collection of skills and resources that teach Claude how to perform tasks in a repeatable, efficient way. Instead of overwhelming your context window, these skills use a smart design to load instructions and code only when they are needed. From automating document creation in Word or Excel to building complex React applications and generative art, these tools unlock new levels of productivity. Explore these custom integrations to master your AI workflow and get more done today. Check out these skills to transform your assistant.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ ddoneapps/fakecall

The Ultimate Fake Call App for Android

Need a discreet way to escape an awkward situation? Check out FakeCall, an open-source Android app that simulates incoming calls so realistically it is nearly impossible to tell the difference. Unlike basic apps that just fake a screen, this one connects directly to your phone's native telecom framework to trigger your real dialer. It features a modern Material 3 interface, lets you schedule calls, add custom audio, and even use automation tools to trigger calls instantly. It is the perfect, polished solution for those moments when you just need a quick exit. Download it and take control today!

πŸ“° https://news.ycombinator.com/item?id=48422001

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ kaimen-inc/co-scientist

Automate Scientific Breakthroughs with AI

Ever wondered if AI could help solve the toughest scientific mysteries? This project is an open source reimplementation of Google's advanced Co-Scientist system, designed to accelerate research by generating novel hypotheses. It uses a multi-agent workflow where specialized AI agents collaborate to propose ideas, verify them for accuracy, and rank them through simulated debates. By using a durable SQLite-backed queue, the system manages complex research goals from start to finish. It is a powerful tool for anyone looking to bridge the gap between creative scientific ideation and structured testing. Explore this intelligent framework to push the boundaries of discovery.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ zdk/lowfat

Slash Your AI Costs with Lowfat

Are you tired of wasting expensive AI tokens on massive command line outputs that simply do not matter? Meet Lowfat, a clever tool designed to strip away digital noise before your AI agent ever sees it. By filtering unnecessary data from your command results, this lightweight utility helps you save significantly on token costs while keeping your workflow clean and efficient. It supports custom plugins, local-first data ownership, and seamless integration with your favorite shell or AI tools. Take control of your AI's context window today and start slimming down your terminal output to make your agents work smarter.

πŸ“° https://news.ycombinator.com/item?id=48409955

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ max-ng/datamoat

Stop Losing Your AI Work History With This Local Backup Tool

Are you worried about your valuable AI work disappearing from cloud histories forever? DataMoat is a powerful local memory archive that lets you export, back up, and search your entire history from tools like ChatGPT, Claude, Cursor, and DeepSeek. By saving your sessions, attachments, and complex skill folders into a normalized, encrypted local vault, it ensures your prompts and reasoning blocks remain fully searchable and reusable across different platforms. It is an essential way to protect your digital knowledge, keep your data private on your own machine, and build a lasting, portable history of every AI project you create.

πŸ“° https://news.ycombinator.com/item?id=48049321

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ albertalef/rubyshell

Stop Writing Bash Scripts! Use RubyShell Instead

Are you tired of debugging messy shell scripts with complex syntax and cryptic error handling? RubyShell is a powerful library that brings the elegance of Ruby to your terminal tasks. It allows you to run shell commands like native Ruby methods, making it easier to chain commands, manage directory scoping, and handle errors with standard Ruby rescue blocks. With features like built-in parallel execution, automatic output parsing for formats like JSON, and clear environment variable management, it turns complex scripting into readable, maintainable code. Experience the developer-friendly way to automate your system tasks and upgrade your command-line workflow today.

πŸ“° https://news.ycombinator.com/item?id=48423466

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ cloudflare/vinext

Run Next.js Apps on Vite with Vinext

Ever wish you could run your Next.js projects using the power and speed of Vite? Meet Vinext, an experimental project that reimplements the Next.js API surface directly on top of Vite. It allows you to leverage Vite for faster development and easier deployments to platforms like Cloudflare Workers while keeping your existing app structure intact. You can even use its built-in migration tools to automatically scan, configure, and swap your build process without destroying your current setup. It is a bold experiment in AI-driven development designed to give you more flexibility. Give it a look to modernize your workflow.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ boona13/image-extender

AI Image Extender & Game Asset Studio

Ever wish you could expand your favorite images without losing quality? This open-source web tool lets you seamlessly extend any picture in any direction using AI. It uses clever Poisson-blending to make seams invisible, ensuring your edits look natural every time. Beyond simple outpainting, it functions as a mini-studio for developers, helping you generate complete parallax backgrounds, animated sprites, and game tiles from a single prompt. Your API keys stay safely in your browser, keeping your data private while you create. Whether you are building a game or just enhancing photos, this tool turns your creative vision into reality.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ nexmoe/vidbee

The Ultimate Video Downloader You Need

Struggling to keep track of videos across different platforms? Meet VidBee, a powerful, open-source video downloader designed to simplify your media collection. Built with Electron and utilizing the robust yt-dlp engine, it lets you grab videos from over one thousand websites with a clean, modern interface. The standout feature is its intelligent RSS auto-download automation, which fetches new uploads from your favorite creators in the background without any manual effort. With flexible format options and easy queue management, VidBee makes saving content effortless. Upgrade your workflow today and start building your own offline library with complete ease.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ deanpeters/product-manager-skills

Supercharge Your Product Management with AI Agents

Want to level up your product management craft using the power of artificial intelligence? This repository provides a collection of forty-nine battle-tested skills and six command workflows designed to elevate both you and your AI agents. By utilizing frameworks from industry leaders, you can effectively frame problems, validate experiments, and kill bad bets faster. Whether you use Claude, Cursor, or other AI agents, these resources teach you the strategy while your agents execute the technical heavy lifting. Start using these professional-grade tools today to transform your workflow and help others in the industry succeed alongside you.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ skywalker-yqz/affordancevla

AffordanceVLA: A New Way for Robots to Understand Actions

Ever wonder how robots can better understand the world around them to perform complex tasks? Meet AffordanceVLA, a powerful model that changes how robots bridge the gap between vision, language, and physical action. Instead of guessing, it uses structured affordance forecasting to identify what objects to interact with, where to touch them, and how to execute precise 3D movements. By integrating these expert components into a specialized transformer architecture, the model creates a smarter roadmap for robotic motion. It is a massive step forward in making machines more capable and intuitive. Check it out to see the future of robotics.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ midudev/autoskills

Automate Your AI Skill Stack with One Command

Are you tired of manually configuring AI agent skills for every new project? Meet this tool that simplifies everything with just one command. By scanning your project files, it intelligently detects your specific tech stack and automatically installs curated, audited AI skills that actually fit your needs. It prioritizes security by verifying every file against a secure manifest before it touches your code, effectively removing the headache of manual setup. Whether you are working with React, Go, or cloud infrastructure, this solution streamlines your workflow instantly. Give it a try today and take your AI development to the next level.

πŸ†” @hackernewsgithubprojects