📦 thudm/indexcache

Speed Up Sparse Attention Models with IndexCache

Are you struggling with slow performance in long-context models using DeepSeek Sparse Attention? IndexCache is a game-changing solution that drastically accelerates inference by cutting out redundant computations. By discovering that adjacent layers share nearly all their selected tokens, this project introduces a clever way to reuse indices instead of calculating them at every single layer. This simple yet effective patch for SGLang and vLLM eliminates up to seventy-five percent of indexer work, delivering massive speedups for your prefill and decode tasks without sacrificing quality. Unlock faster, more efficient AI performance today with this powerful optimization technique.

📰 https://news.ycombinator.com/item?id=48569165

🆔 @hackernewsgithubprojects

📦 vakovalskii/codbash

Stop Losing Your AI Coding Sessions!

Are you tired of losing track of your AI coding sessions scattered across different tools and folders? Meet Codbash, a powerful browser-based control room designed to bring all your AI coding history into one organized place. Whether you are using Claude Code, Cursor, Codex, or other popular agents, this dashboard lets you search, resume, and manage your workflows with ease. You can view session replays, monitor live statuses, and even calculate costs based on real token usage. With its intuitive interface, you can finally stop digging through logs and start coding more efficiently. Simplify your AI development workflow today.

🆔 @hackernewsgithubprojects

📦 bvisible/mcp-ssh-manager

Control Your Servers with AI: MCP SSH Manager

Are you tired of manually jumping between servers to manage your infrastructure? Meet the MCP SSH Manager, a powerful tool designed to bridge the gap between your AI assistant and your remote machines. By using the Model Context Protocol, this software empowers Claude Code and OpenAI Codex to securely execute commands, transfer files, and monitor health across all your servers. With thirty-seven built-in tools covering everything from automated backups to database management, it turns complex DevOps tasks into simple natural language requests. Supercharge your remote management workflow and let your AI handle the heavy lifting starting today.

🆔 @hackernewsgithubprojects

📦 greensock/gsap-skills

Master GSAP Animations with AI: Official Skills for Developers

Struggling to get your web animations just right? The official GSAP skills repository is a game changer for developers using AI coding agents. It provides a curated collection of specialized instructions that teach AI exactly how to use the GreenSock Animation Platform correctly. From core tweening methods and performance optimization to framework-specific integration for React, Vue, and Svelte, these skills ensure your generated code follows industry best practices. Stop wasting time debugging animation patterns and let these precise guidelines handle the complexity for you. Integrate these skills today to streamline your workflow and start building smoother, professional-grade animations effortlessly.

🆔 @hackernewsgithubprojects

📦 caramaschihg/awesome-ai-agents-2026

The Ultimate Guide to AI Agents in 2026

Are you feeling overwhelmed by the endless wave of new artificial intelligence tools? This repository is your ultimate survival guide, acting as a massive, monthly-updated directory that organizes over three hundred and forty of the best AI agents, frameworks, and productivity tools into twenty clear categories. Whether you need autonomous coding assistants like Devin, multi-agent frameworks to scale your workflow, or creative tools for media generation, this project cuts through the noise. It helps you quickly identify the right technology for your specific needs, making it an essential resource for anyone looking to master the agentic future of software.

🆔 @hackernewsgithubprojects

📦 aaronsb/obsidian-mcp-plugin

Give Your AI Vault-Level Intelligence with Obsidian MCP

Unlock the full potential of your knowledge base by giving AI direct agency over your Obsidian notes. This repository provides a high-performance server that connects your local vault to AI assistants using the open Model Context Protocol. Instead of seeing isolated files, your AI can now traverse your knowledge graph, perform semantic searches, and execute complex file operations like splitting or merging notes. It seamlessly bridges the gap between your private data and powerful AI tools, enabling you to synthesize ideas and explore connections effortlessly. It is a game changer for turning your notes into a truly intelligent research assistant.

🆔 @hackernewsgithubprojects

📦 modelscope/twinkle

Supercharge Your Model Training with Twinkle

Struggling to get your AI models performing at their peak? Meet Twinkle, a comprehensive training workbench designed to help your models shine. This project provides a powerful infrastructure for fine-tuning and optimizing large language models, offering modular components like custom loss functions, advanced samplers, and specialized checkpoint engines. Whether you are dealing with complex data loading or need robust NPU support, it streamlines the training process through a flexible, scalable architecture. It is essentially a professional-grade toolkit that simplifies sophisticated training workflows, allowing you to focus on achieving better results. Dive into this repository to elevate your model development journey today.

🆔 @hackernewsgithubprojects

📦 ballred/obsidian-claude-pkm

Stop Failing Your Goals: Obsidian + Claude AI Accountability

Struggling to keep your goals and daily tasks in sync? This powerful starter kit transforms your Obsidian vault into a high-performance accountability system using Claude Code. By connecting your three-year vision directly to your daily actions, it uses AI-powered agents and slash commands to ensure nothing ever falls through the cracks. Whether you are automating weekly reviews, tracking project progress, or aligning daily habits with long-term milestones, this system keeps you focused and moving forward. Stop guessing where your time goes and start building a workflow that keeps you accountable to your own success starting today.

🆔 @hackernewsgithubprojects

📦 guria/nehir

Take Control of Your Desktop with Nehir

Ever wish your Mac worked more like a professional tiling window manager? Meet Nehir, a scrolling tiling window manager for macOS that transforms how you organize your digital workspace. Inspired by the Niri layout paradigm, it automatically arranges your windows into columns that flow horizontally across your screen. With features like configurable colored window borders, per-monitor status bars, and powerful keyboard-driven navigation, it is designed to keep your workflow efficient and fluid. Whether you are using app rules or custom hotkeys, Nehir gives you total control over your screen real estate. Level up your productivity and streamline your desktop flow today.

🆔 @hackernewsgithubprojects

📦 codehamr/codehamr

Meet the Minimalist Coding Agent for Your Terminal

Are you tired of complex coding tools that clutter your workflow? Meet a minimal, local-first coding agent designed specifically for your terminal. It prioritizes simplicity by avoiding unnecessary features, focusing instead on a streamlined loop that calls tools like bash, file readers, and editors to get the job done efficiently. Built to work with local models or OpenAI-compatible endpoints, it lets you maintain full control over your context window. By automating tasks directly in your project environment, it ensures your code stays clean and manageable. Experience a lighter way to code and let your terminal do the heavy lifting.

🆔 @hackernewsgithubprojects

📦 lobstertrap/lola

Stop configuring AI assistants one by one: Meet Lola

Tired of setting up the same AI skills across different assistants? Meet Lola, the universal AI package manager that solves this fragmentation. Just like a package manager for your operating system, Lola allows you to write your AI context modules and skills once and deploy them anywhere. Whether you use Claude Code, Cursor, Copilot, or Gemini, you can manage your tools with simple commands. It handles the conversion, pathing, and installation automatically, letting you spend less time configuring and more time building. Start creating your own portable AI modules today and experience a truly unified workflow across all your favorite coding tools.

📰 https://news.ycombinator.com/item?id=47933332

🆔 @hackernewsgithubprojects

📦 fireredteam/firered-openstoryline

AI Video Editing Made Simple: Meet FireRed-OpenStoryline

Tired of spending hours on manual video editing? Meet FireRed-OpenStoryline, an innovative AI agent that turns complex video creation into a simple, intention-driven conversation. Instead of fighting with timelines, you use natural language to direct the entire process. It uses LLM-powered planning to handle tasks like smart media searching, automatic script generation, and even AI-driven transitions. You can save your workflows as reusable skills, ensuring consistent storytelling for all your future projects. Whether you are a beginner or a creator looking for efficiency, this tool lets you focus on your vision. Start directing your stories with AI today.

🆔 @hackernewsgithubprojects

📦 llmquant/awesome-trading-agents

Build Your Own AI Trading Agent

Are you ready to automate your market research and trading strategy? This repository is a goldmine for anyone looking to build or use LLM-driven trading agents. It provides a curated collection of agents, market data servers, and trading skills specifically designed to help you analyze markets and execute trades. Whether you need an autonomous hedge fund setup or simple tools to pull live financial data, this list offers a clear starting point for agentic finance. Dive into these building blocks today to start crafting your own intelligent trading systems and take your market analysis to the next level.

🆔 @hackernewsgithubprojects

📦 macton/nagent

Stop Building Agents: Try This Data-Oriented Approach

Stop chasing the idea of AI agents and start focusing on your data. The Nagent project is a reference implementation that treats AI workflows as transparent, data-oriented processes rather than opaque, black-box loops. Instead of relying on hidden memory, Nagent stores everything in editable text files on your disk. This allows you to inspect, branch, and maintain your conversation history just like any other source code. By keeping your prompts, tool outputs, and context in durable files, you turn fleeting LLM interactions into permanent, shareable project knowledge. Stop patching bad output and start owning your data today.

📰 https://news.ycombinator.com/item?id=48432410

🆔 @hackernewsgithubprojects

📦 anthonysturdy/micro-radar

Build Your Own Miniature Desktop Flight Radar

Ever wanted to track live air traffic right from your desk? The Micro Radar project lets you build a tiny, dedicated hardware device that visualizes nearby aircraft in real time. Using an ESP32-C3 module and a compact display, this open-source project pulls flight data from the OpenSky network to animate planes moving across a custom circular interface. You will need to 3D print an enclosure and assemble the electronics, but the result is a functional, elegant aviation tracker for your workspace. Grab your soldering iron and some basic components to start monitoring the skies above your home today.

📰 https://news.ycombinator.com/item?id=48547732

🆔 @hackernewsgithubprojects

📦 917dhj/deeppapernote

Stop Manually Summarizing Papers: Use DeepPaperNote

Do you find it exhausting to manually turn complex academic papers into useful, organized notes? DeepPaperNote is an agent skill that solves this by automating the heavy lifting of your research workflow. It integrates with tools like Claude Code and Codex to ingest papers, extract key mechanisms, and generate high-quality, Obsidian-ready Markdown notes. Instead of jumping between files, you get structured summaries that prioritize technical details, equations, and clear figure contexts. It effectively handles the mechanical work of note production so you can focus on mastering the actual content. Streamline your study process and build a better knowledge base today.

🆔 @hackernewsgithubprojects

📦 lyra-stellai/web-browser-agent

Build Your Own Knowledge Graph from the Web!

Struggling to keep track of everything you read online? This lightweight web application turns your browsing into a personal knowledge base. It lets you search the web, summarize articles with AI, and highlight key points to save as structured notes. The coolest part is its ability to integrate these chunks into a persistent knowledge graph, using LLMs to automatically extract entities and connect related ideas over time. Whether you use popular AI models or a local fallback, you can finally turn scattered web content into a connected, searchable brain. Start building your own intelligent knowledge network today!

🆔 @hackernewsgithubprojects

📦 dbwls99706/ros2-engineering-skills

Supercharge Your ROS 2 Development with AI Agents

Want to build professional-grade robotics software without getting lost in complex configurations? This project provides a powerful engineering skill module designed to give AI coding agents deep expertise in ROS 2. Instead of generic advice, you get precise decision frameworks for production challenges like QoS profile tuning, lifecycle node management, and hardware interfacing with ros2_control. It uses a progressive disclosure structure to keep your AI assistant focused and efficient, whether you are setting up workspaces, navigating Nav2, or deploying multi-robot fleets. Stop guessing and start coding with reliable, distro-aware patterns that ensure your robot projects are truly production-ready.

🆔 @hackernewsgithubprojects

📦 boogu-project/boogu-image

Meet Boogu-Image: High-Performance Open-Source AI Generation

Ever wonder how to get professional-grade image generation without the massive data requirements of closed systems? Meet Boogu-Image, an impressive open-source model family designed for high-quality text-to-image generation and precise image editing. By refining training pipelines and data quality, it delivers near-closed-source performance using significantly less training data than typical models. Whether you need stable bilingual text rendering, photorealistic visuals, or flexible editing tools, this model provides powerful capabilities in a compact package. Explore their diverse variants like the fast Turbo mode or the Base model and start creating high-quality AI visuals on your own local hardware today.

🆔 @hackernewsgithubprojects

📦 openhands/agent-canvas

Build Smarter with OpenHands Agent Canvas

Tired of repetitive coding tasks slowing you down? Meet OpenHands Agent Canvas, a powerful self-hostable platform designed to automate your development workflow right from your local machine. This project lets you connect your own tools and build intelligent agents to handle complex coding projects efficiently. With features like built-in terminal access, file system integration, and advanced task tracking, it transforms your browser into a command center for AI-assisted development. Whether you are managing backend services or launching custom automation, this platform gives you full control. Take charge of your coding process and start building your own automated workspace today.

🆔 @hackernewsgithubprojects