📦 4thfever/cultivation-world-simulator

Build Your Own AI Cultivation World Simulator

Ever wanted to play as a god in your own martial arts world? This open source Cultivation World Simulator lets you do exactly that by turning every character into an autonomous AI agent. Powered by large language models, this project creates a dynamic, emerging narrative where NPCs form relationships, build sects, and make independent choices based on a robust set of underlying cultivation rules. Since there are no pre-scripted events, you get to observe a living, breathing world evolve in real time. It is a fascinating way to watch epic stories unfold through pure simulation, so start your own journey today.

📰 https://news.ycombinator.com/item?id=48552548

🆔 @hackernewsgithubprojects

📦 murrol/stablemotion

Clean Up Motion Data Without Clean Datasets!

Have you ever struggled with noisy or corrupted motion capture data that just won't cooperate? StableMotion is a groundbreaking solution presented at SIGGRAPH Asia 2025 that changes the game by letting you train motion cleanup models using only unpaired, corrupted data. You no longer need to rely on handcrafted pairs or synthetic augmentations to fix your animations. This project provides a complete toolkit for preprocessing datasets, training custom models, and running inference to restore your motion sequences to a clean state. It is an incredibly powerful resource for anyone working with raw mocap, so start cleaning your animations today!

🆔 @hackernewsgithubprojects

📦 huangserva/3dcellforge

Build and Inspect 3D Models Effortlessly

Ever wanted to turn static images into interactive 3D experiences? This powerful studio lets you generate, inspect, and present 3D models right in your browser. Built with React and Three.js, it offers a professional workbench to transform reference images or import GLB files into a polished 3D scene. You get full control with orbit navigation, cinematic camera paths for demos, and even quality scoring for your assets. Whether you are analyzing complex structures or showcasing digital specimens, it simplifies your workflow with an intuitive library and powerful export tools. Dive in and start creating your own interactive 3D collection today.

🆔 @hackernewsgithubprojects

📦 labarba/sciwrite

Improve Your Scientific Writing Instantly with AI

Struggling to make your scientific manuscript sound professional and clear? This repository offers a powerful Agent Skill designed to act as your personal scientific writing coach. By encoding the proven editorial methodology from Stanford professor Dr. Kristin Sainani, it transforms cluttered academic prose into precise communication. The skill performs five systematic audit passes, ranging from stripping away filler phrases and fixing passive voice to ensuring keyword consistency across your entire paper. It works across major AI platforms, helping you refine your delivery without ever altering your actual research data. Give your writing the polish it deserves today.

🆔 @hackernewsgithubprojects

📦 jeidoban/ironsmith

Build Custom Mac Apps Instantly With AI!

Ever wish you had a custom tool to handle your specific Mac tasks, but lack the coding skills to build one? Meet Ironsmith, a clever menu bar app that turns your ideas into reality using the power of AI. Simply describe what you need, and Ironsmith writes, builds, and packages a native SwiftUI application for you in minutes. It supports both cloud and local models, ensuring you have the flexibility to create everything from simple file converters to unique utilities. Stop hunting for apps that almost do what you want and start building your own perfect workflow today!

🆔 @hackernewsgithubprojects

📦 homeassistant-ai/ha-mcp

Control Home Assistant with AI: The Power of MCP

Imagine talking to your smart home and actually having it listen, learn, and adapt to your needs without complex manual menus. This project is a comprehensive Model Context Protocol server that bridges your AI assistant directly to your Home Assistant setup. By using natural language, you can command your devices, build automations, or even debug complex scripts through simple chat. Unlike basic integrations, this tool grants your AI deep access to manage your dashboards, history, and system configurations. It turns your assistant into a true smart home architect, helping you spend less time configuring and more time enjoying home.

📰 https://news.ycombinator.com/item?id=48553124

🆔 @hackernewsgithubprojects

📦 jd-opensource/joyai-vl-interaction

Meet JoyAI: The Future of Real-Time AI Interaction

Are you tired of waiting for AI to respond to your commands? Standard AI assistants are trapped in a turn-based loop, only speaking when spoken to, but the real world never stops. Enter JoyAI-VL-Interaction, a groundbreaking, fully open-source vision-language model that brings true real-time presence to your devices. Unlike static models, this 8B-scale system watches your environment continuously, deciding on its own when to speak or stay silent. By using advanced predictive codecs and massive time-aligned training data, it offers proactive, instant assistance. Explore the full stack and start building your own always-present, intelligent assistant today to experience a new era.

🆔 @hackernewsgithubprojects

📦 ottermind/nubase

Turn AI-Generated Code Into Real Apps With Nubase

Want to turn your AI-written code into a fully functional application without the backend headache? Meet Nubase, an open-source, AI-native platform designed for modern product teams and agentic applications. It solves the complexity of infrastructure by bundling memory, database, storage, and authentication into a single, self-hostable service. Whether you are building complex AI agents or rapid prototypes, Nubase provides the essential building blocks to get your projects production-ready fast. It simplifies the transition from AI concepts to actual software by providing a unified backend layer. Give your AI-generated code the solid foundation it deserves and start building smarter today.

📰 https://news.ycombinator.com/item?id=48552581

🆔 @hackernewsgithubprojects

📦 v4bel/itscape

ITScape: The First Guest-to-Host KVM/arm64 Escape Vulnerability

Imagine a security flaw so deep it lets a virtual machine break out of its cage and take full control of the host computer. That is exactly what ITScape is. This newly discovered vulnerability targets KVM on arm64 processors, specifically within the interrupt translation system. It allows a malicious guest to bypass isolation barriers and execute commands on the host kernel with root privileges. By exploiting a race condition in memory management, it creates a way to rewrite host memory directly. It is a major wake-up call for cloud infrastructure security, reminding us to keep systems fully patched and protected.

📰 https://news.ycombinator.com/item?id=48491357

🆔 @hackernewsgithubprojects

📦 telepat-io/otto

Automate Web Tasks on Real Browser Tabs with Otto

Are you tired of managing complex, resource-heavy headless browser farms just to automate simple web workflows? Meet Otto, a powerful platform that lets you control real Chrome tabs directly from your terminal or scripts. Instead of setting up clunky infrastructure, you simply send commands to a lightweight relay that communicates with a browser extension. It handles everything from clicks and text extraction to network interception with total precision. Whether you are building AI agents or streamlining CI testing, Otto keeps your workflows secure, fast, and deterministic. Take back control of your automation stack and start running smarter today.

🆔 @hackernewsgithubprojects

📦 karimknaebel/surge

SurGe: Elevate Your 3D Geometry Reconstruction

Want to transform flat images into detailed 3D geometry? Meet SurGe, a powerful tool designed to improve surface geometry in point maps. Whether you are working with individual images or entire directories, SurGe allows you to generate precise meshes, point clouds, and depth maps with ease. It features a versatile Python API for developers and a user-friendly Gradio app for quick, interactive testing. With its specialized neighborhood attention decoder, you can achieve high-quality reconstructions effortlessly. From architectural visualization to creative 3D projects, SurGe helps you capture the scene with accuracy. Give your data the depth it deserves today.

🆔 @hackernewsgithubprojects

📦 macrodata-labs/refiner

Scale Your Machine Learning Data with Refiner

Struggling to prepare massive datasets for your machine learning models? Meet Refiner, an open-source framework specifically designed to streamline the processing of complex robotics and multimodal data. Instead of relying on generic tools, Refiner provides specialized pipeline primitives that allow you to efficiently clean, annotate, and transform large-scale data with ease. You can develop your processing logic locally and then scale it seamlessly across your infrastructure. Whether you are trimming motion data or running model-based inference, this framework simplifies your workflow by handling the heavy lifting of sharding and distributed execution so you can focus on building better models.

🆔 @hackernewsgithubprojects

📦 microsoft/fara

Microsoft Fara-7B: The Small AI Agent That Controls Your PC

Have you ever wanted an AI that can actually use your computer like a human? Meet Fara-7B, an ultra-compact, seven-billion-parameter agent designed by Microsoft specifically for computer use. Unlike standard chatbots, Fara-7B perceives your screen through direct screenshots and uses your mouse and keyboard to handle complex, multi-step web tasks. By running locally on your device, it ensures your data stays private while delivering fast, state-of-the-art performance on benchmarks like WebVoyager. Whether it is shopping, booking travel, or filling out forms, Fara-7B transforms your PC into a truly automated workspace. Experience the future of local, efficient, and private AI automation.

🆔 @hackernewsgithubprojects

📦 gary23w/neuron-db

Stop Overloading Your AI: Use This Brain-Like Memory

Ever feel like your AI forgets everything or runs out of space? Meet a clever way to give your LLM a massive, brain-like memory that lives outside its limited context window. This project provides an associative memory system that lets you save facts in plain language and recall them by meaning without needing complex databases, embeddings, or expensive model training. It runs entirely on your machine, using a fast Rust core that handles millions of facts with microsecond latency. It is basically a searchable, long-term storage layer that keeps your AI smart, efficient, and perfectly accurate at any scale.

📰 https://news.ycombinator.com/item?id=48527346

🆔 @hackernewsgithubprojects

📦 thu-team-eureka/eurekagent

Automate Scientific Research with EurekAgent

Ever wonder how to automate complex scientific discovery? Meet EurekAgent, an autonomous research system designed to tackle metric-driven tasks through a continuous experimentation loop. By defining your problem and evaluation criteria, the system coordinates specialized agents to propose diverse approaches, implement code, and run experiments in isolated Docker containers. It handles the entire research cycle—iterating toward breakthroughs while tracking your progress and costs in real time. Whether you are working in mathematics or machine learning, this tool simplifies discovery by managing the heavy lifting for you. Give your research a massive boost and start automating your experiments today.

🆔 @hackernewsgithubprojects

📦 witcherofresearch/qwen-image-style-transfer

Master Content-Preserving Style Transfer with QwenStyle

Ever wondered how to perfectly apply an artistic style to your photos while keeping the subject exactly as it is? Meet QwenStyle, the first model that brings high-efficiency content-preserving style transfer to the Qwen-Image-Edit framework. By leveraging specific style and lightning lora models, this project allows you to transform images in just four sampling steps without losing key visual details. Whether you are aiming for rapid stylized edits or experimenting with custom content, this repository provides the raw code and diffusion-based models you need to get started. Dive into this efficient, creative tool to transform your digital imagery today.

🆔 @hackernewsgithubprojects

📦 ai-hypercomputer/accelerator-agents

Supercharge Your TPU Development with Accelerator Agents

Optimizing machine learning models for high-performance hardware is a massive challenge, but Accelerator Agents makes it easier by providing specialized AI tools designed for Google Cloud TPUs. This project uses Gemini-powered agents to automate complex tasks like migrating PyTorch code to JAX and optimizing custom Pallas kernels for maximum performance. With built-in tools like MaxCode for code migration and JAXBench for performance evaluation, developers can streamline their workflow and unlock better hardware utilization. Whether you are debugging kernels or porting models, these agents provide the intelligent assistance you need to build faster today.

🆔 @hackernewsgithubprojects

📦 eit-nlp/awesome-streaming-llms

Understanding Streaming Large Language Models

Ever wondered how AI models handle live, incoming information? This repository is a goldmine for anyone tracking the evolution of streaming large language models and multimodal systems. It provides a comprehensive survey that helps clear up confusion by categorizing these models into three progressive paradigms: output streaming, sequential streaming, and concurrent streaming. Whether it is managing long-term memory or deciding exactly when to read and respond, this collection covers the essential research shaping real-time AI interaction. Dive into this structured resource to explore the cutting edge of how machines bridge the gap from static inference to full-duplex dynamic interaction.

🆔 @hackernewsgithubprojects

📦 ndvbd/gefen

Train Larger AI Models with Gefen

Struggling with GPU memory limits while training massive AI models? Meet Gefen, a clever drop-in replacement for the standard AdamW optimizer designed for high efficiency. By utilizing specialized fused CUDA kernels, Gefen slashes your optimizer state memory footprint by up to eight times without sacrificing performance. This gives you the freedom to train significantly larger models or utilize much bigger batch sizes than before. It is incredibly easy to implement, requiring you to change just two lines of code in your existing workflow. Optimize your training pipeline today and start pushing your hardware to its full potential.

🆔 @hackernewsgithubprojects

📦 supermemoryai/claude-supermemory

Give Your Claude Code Persistent Memory

Ever wish your AI coding assistant remembered your previous sessions or project decisions? The Claude Supermemory plugin changes that by giving Claude persistent, long-term memory. It automatically captures important conversations and allows you to manually save key architectural decisions or bug fixes for the whole team to reference later. With smart, reasoned recall, Claude only searches your past memories when it actually helps your current task. Whether you are tracking coding styles or complex project patterns, this tool helps your AI grow alongside you. Install the plugin today to stop repeating yourself and start building smarter.

🆔 @hackernewsgithubprojects