📦 wbopan/retro-harness

Supercharge Your AI Agents Without Human Labels

Want to make your AI agents smarter without needing expensive human-labeled data? Meet Retrospective Harness Optimization, or RHO. Instead of relying on external feedback, RHO allows agents to improve their performance by analyzing their own past experiences. It intelligently selects diverse tasks, performs parallel rollouts, and uses self-consistency checks to refine its own skills and tools. In practice, this approach can dramatically boost success rates on challenging benchmarks by learning directly from unlabeled trajectories. It is a powerful way to automate agent improvement, so start evolving your agent's capabilities today and see what it can achieve.

🆔 @hackernewsgithubprojects

📦 huixiai/rhinovla

RhinoVLA: Real-Time Robot Control on Edge AI Chips

Imagine robots that think and react in real-time, all running directly on the hardware. RhinoVLA is a groundbreaking cross-embodiment system designed for edge-side robot control. By combining a powerful model with the efficiency of the R1 chip, it achieves an impressive inference speed that meets the gold standard for closed-loop control. It uses clever techniques like instance adaptation to work seamlessly across different robot platforms, from instruction following to complex bimanual tasks. This project bridges the gap between sophisticated artificial intelligence and physical robotics, proving that real-time, high-performance edge computing is the future of intelligent autonomous systems.

🆔 @hackernewsgithubprojects

📦 5uck1ess/tts-bench

Stop Guessing: How to Benchmark Your AI Voice Models

Are you struggling to find the best local text to speech model for your specific hardware? This repository provides a comprehensive benchmarking suite designed to evaluate speed, quality, and voice cloning capabilities across diverse platforms including Windows, Linux, and Mac. By measuring cold and warm latency, memory usage, and real-time factors, it helps you identify the most efficient models for your system. It even includes a blind voting arena where you can compare audio clips to find the most natural sounding voices. Stop guessing and start testing your models today to get the performance and quality you truly need.

🆔 @hackernewsgithubprojects

📦 luke8086/gentleos32

Build Your Own OS for Vintage PCs

Ever wanted to breathe new life into an old 32-bit PC? Meet GentleOS, a minimalist hobby operating system designed specifically for vintage hardware. This project provides a clean, monolithic platform that runs directly on bare metal, requiring only an i386 processor and a little RAM. It handles core essentials like VGA graphics, keyboard input, and even a functional mouse interface to support interactive graphical apps. Whether you are tinkering with retro setups or just curious about kernel development, this open-source project offers a fascinating, hands-on way to explore how operating systems talk to hardware. Dive in and start building today.

📰 https://news.ycombinator.com/item?id=48458890

🆔 @hackernewsgithubprojects

📦 cporter202/api-mega-list

The Ultimate API Treasure Trove for Your Next Project

Are you tired of hunting for the right tools to fuel your latest app idea? This repository is a goldmine for developers and creators, offering a massive collection of ready-to-use APIs designed to power everything from simple automations to complex applications. Whether you need to extract YouTube transcripts, automate job data collection from platforms like LinkedIn, or integrate advanced AI assistants using the Model Context Protocol, this list has you covered. It solves the headache of searching for reliable services by gathering high-quality, practical tools in one place. Explore this library today to streamline your workflow and start building faster.

🆔 @hackernewsgithubprojects

📦 stability-ai/stable-audio-3

Meet Stable Audio 3: The Future of AI Music Generation

Ever wanted to turn simple text into high-quality studio audio? Stable Audio 3 is an open platform designed to make that dream a reality. This tool uses advanced diffusion technology to generate music and sound effects in seconds, allowing you to edit existing recordings or even expand clips. Whether you are creating beats or soundscapes, it offers impressive flexibility and high fidelity. It even supports personalized style training using LoRA, so you can tailor the output to your creative vision. It is built for speed and efficiency, making professional-grade audio generation accessible to everyone. Ready to transform your creative workflow?

🆔 @hackernewsgithubprojects

📦 leochlon/ntkmirror

Train LLMs Without LoRA: The NTK-Mirror Way

Want to fine-tune your language model without the overhead of LoRA? This repository offers a clever alternative for Hugging Face causal language models. Instead of permanent weight edits, it learns a small, sparse controller consisting of signed log-gates on the decoder layer channels. You can fit these controllers on your own data, compose multiple task-specific controllers by adding their log-gate values, and even store them as persistent memory for specific styles or procedures. It is a unique way to steer model behavior without changing the base weights. Explore this approach to see how light and modular your model fine-tuning can actually become.

🆔 @hackernewsgithubprojects

📦 patriceckhart/zot

Build Smarter with Zot: A Lightweight Coding Agent Harness

Stop juggling clunky tools and meet Zot, a lightweight coding agent harness built entirely in Go. Designed to streamline your development, Zot functions as a flexible bridge to dozens of top AI models like Anthropic, OpenAI, and local LLMs through Ollama. It provides essential tools for reading, writing, and editing files, plus bash command execution, all within an intuitive terminal interface. Whether you need an interactive chat experience or automation via JSON-RPC, Zot adapts to your workflow. With features like session management, background subagents for parallel tasks, and custom extensions, it is the perfect companion to supercharge your coding productivity.

📰 https://news.ycombinator.com/item?id=48465356

🆔 @hackernewsgithubprojects

📦 ernie-research/nava

NAVA: Next-Level Audio-Visual Generation

Ever wondered how to make video and audio perfectly sync without tedious post-editing? Meet NAVA, a cutting-edge framework that generates audio and video simultaneously with native alignment. Instead of just creating visuals, NAVA bridges the gap by building a dedicated alignment space for both modalities, ensuring your sounds match every movement on screen. It even supports multi-timbre voice control, letting you bind specific voices to characters for incredible realism. Whether you are generating high-quality seven-twenty-p video or complex audio, this open-source tool offers end-to-end control for your creative projects. Dive into the NAVA repository today to revolutionize your generation workflow.

🆔 @hackernewsgithubprojects

📦 trystan-sa/rproc

A Windows 11 Style Task Manager for Linux

Ever wish your Linux system had a task manager as clean and functional as the one in Windows 11? Meet this project, a modern, efficient resource and process monitor built entirely in Rust. It provides live, easy-to-read charts for your CPU, memory, disk, network, and GPU, while letting you effortlessly sort, filter, and manage running processes. What makes it stand out is its commitment to being incredibly memory-frugal, offering a smooth experience without the heavy resource footprint of typical monitors. Whether you need to track services or startup entries, this tool delivers a polished, high-performance interface for total system control.

🆔 @hackernewsgithubprojects

📦 zai-org/scail-2

Take Control of AI Character Animation with SCAIL-2

Tired of AI character animations that look robotic or struggle with complex movements? Meet SCAIL-2, an open-source model designed to unify controlled character animation through end-to-end in-context conditioning. By bypassing rigid intermediate pose representations, this project solves the limitations of previous methods, allowing for more natural motion across diverse subjects, including animals and cross-identity replacements. Whether you need to animate a specific character or swap people in existing videos, SCAIL-2 uses advanced motion transfer interfaces to deliver high-quality, flexible results. Dive into the repository to start crafting your own seamless, professional-grade animations today and push your creative projects further.

🆔 @hackernewsgithubprojects

📦 futuremls-lab/oscar

Optimize LLM Performance with OSCAR

Struggling with high memory usage in your large language models? Meet OSCAR, a technique designed to tackle the bottleneck of key-value cache memory. By using offline spectral covariance-aware rotation, this project enables efficient two-bit quantization for your KV cache, significantly reducing memory footprints without sacrificing performance. It integrates directly into advanced serving frameworks, offering robust support for various hardware platforms and complex model architectures. Whether you are scaling up inference or optimizing memory for resource-constrained environments, OSCAR provides a powerful path toward faster and more efficient deployments. Dive into this repository to explore how you can streamline your AI infrastructure today.

🆔 @hackernewsgithubprojects

📦 trystan-sa/claude-design-system-prompt

Stop Building AI Slop: The Pro Designer's Prompt

Are you tired of your AI-generated designs looking like generic templates with aggressive gradients and overuse of emojis? This repository is a game-changer for anyone who wants to create polished, intentional work. It provides a reverse-engineered system prompt and a library of fourteen procedural skills that turn an LLM into an opinionated, accessibility-aware design collaborator. By following a disciplined philosophy—from visual hierarchy and content rhythm to semantic HTML—this project helps you move beyond AI tropes. Whether you need to extract components or audit accessibility, you can chain these skills together to build high-quality, professional prototypes that truly stand out.

🆔 @hackernewsgithubprojects

📦 memovai/mimiclaw

Build a Pocket AI Assistant on a $5 Chip

Ever wanted your own personal AI assistant that fits in the palm of your hand? Forget complex servers or Linux—you can now turn a tiny five-dollar chip into a powerful, always-on agent. By leveraging pure C code on an ESP32-S3 development board, this project creates a hardware-based AI that connects to your WiFi and interacts directly through Telegram. It supports advanced models like Claude and GPT, handles web searches, and even learns through long-term memory stored locally on flash. It is a brilliant, lightweight way to bring autonomous intelligence into your physical space. Start building your own AI companion today.

🆔 @hackernewsgithubprojects

📦 allenai/molmoact2

Meet MolmoAct2: The Future of Robot Action Reasoning

Imagine a robot that can actually reason about its environment to perform complex tasks. MolmoAct2 is a powerful open family of action reasoning models designed for real-world robot control. By combining an embodied vision-language backbone with a flow-matching continuous action expert, these models enable robots to handle intricate manipulation tasks effectively. The repository provides everything you need to get started, including base foundation checkpoints, fine-tuned policies for popular robot platforms, and specialized inference servers for deployment. Whether you are working with Franka or bimanual setups, MolmoAct2 bridges the gap between high-level reasoning and physical movement. Start exploring these models today.

🆔 @hackernewsgithubprojects

📦 colemurray/claude-code-otel

Master Your AI Development Costs with Claude Code Observability

Are you curious about exactly how much your AI coding sessions are costing you? The Claude Code Observability Stack is a powerful toolkit designed to give you deep insights into your AI-assisted development workflows. It uses OpenTelemetry to track critical data like API costs, token usage, tool performance, and overall developer productivity. With ready-to-use Grafana dashboards, you can monitor real-time metrics, identify performance bottlenecks, and visualize your spending trends with ease. It is the perfect way to gain full visibility into your AI tool investment, so stop guessing about your usage and start optimizing your development efficiency today.

🆔 @hackernewsgithubprojects

📦 tjboudreaux/cc-thinking-skills

Supercharge Your AI Reasoning with 39 Mental Models

Stop relying on ad-hoc guesses and start using structured logic to solve your hardest coding challenges. This collection introduces 39 essential mental models and critical thinking frameworks, including first principles, Bayesian reasoning, and the OODA loop, designed specifically to enhance how your AI agent processes complex decisions. By invoking these skills, like the model router or scientific method, you provide a clear roadmap for better debugging and strategic planning. The best part is the commitment to total transparency, with every framework backed by a rigorous evaluation pipeline that prioritizes honest results over hype. Start thinking clearer and more effectively today.

🆔 @hackernewsgithubprojects

📦 guysandler/canvasrefined

Take Control of Your Canvas Experience with Canvas Refined

Are you looking to improve your Canvas interface? Meet Canvas Refined, an open source browser extension designed to enhance your dashboard and productivity. Tired of clunky layouts, this project brings you a fully customizable dark mode with automatic scheduling, a helpful GPA calculator for high school and college, and an upgraded to-do list to keep you on track. It also features dashboard cards with custom color palettes and assignment previews, all while keeping your data private and local on your machine. Take back your study space and customize your learning experience today with this simple, community-driven tool.

🆔 @hackernewsgithubprojects

📦 hehee9/2026-csat

AI vs The Korean SAT: Can Machines Ace the Test?

Ever wonder how today’s advanced AI models handle the toughest academic exams? This project provides a fascinating look into the performance of various large language models on the 2026 Korean SAT. The repository features a dedicated dashboard built with React that allows users to visualize and compare how different models navigate complex test questions across multiple subjects. By organizing structured data and performance metrics, it creates an accessible way to analyze machine reasoning and accuracy. It is a brilliant example of tracking AI capabilities in real-time academic settings, offering a clear view of where these technologies currently stand.

🆔 @hackernewsgithubprojects

📦 deepfates/cantrip

Cantrip: The Elixir Agent Runtime You Need to Know

Ever wondered how to make your code speak for itself? Meet Cantrip, an extensible, self-documenting multi-agent environment built in Elixir. It treats language models as active participants, wrapping them in reusable circles that define how they interact with your world. Whether you are automating codebase analysis or managing persistent entities, Cantrip lets you design specialized agents that reason, delegate tasks, and record their progress in a durable loom. By leveraging Elixir’s robust concurrency, it turns simple prompts into complex, reliable workflows. Start building smarter, more autonomous systems today, and discover how powerful your development loop can become.

🆔 @hackernewsgithubprojects

📦 assafwoo/homebrew-pandafilter

Stop AI Agents from Wasting Your Context Window

Are your AI coding agents wasting tokens on useless noise? Meet PandaFilter, the intelligence layer that sits between your tools and your AI. It intercepts command output to strip away fluff like progress bars and module graphs while preserving only what matters. Using BERT-powered compression and smart routing, it handles everything from test failures to file reads, often slashing token usage by over ninety percent. It works across major agents like Claude Code, Cursor, and Windsurf to keep your sessions clean and your costs low. Install it today to let your AI agent think faster and stay focused.

📰 https://news.ycombinator.com/item?id=47814614

🆔 @hackernewsgithubprojects