📦 miromindai/mirothinker

Meet MiroThinker: The Next Level in Deep Research AI

Are you ready for an AI agent that actually masters deep research and complex prediction? Meet MiroThinker, a powerful open-source research framework designed to handle long-horizon reasoning and intricate multi-step tasks. It solves the difficulty of reliable agentic workflows by supporting massive context windows and hundreds of automated tool interactions. Whether you need deep analysis or structured web research, MiroThinker provides the tools and workflows to get the job done. With its state-of-the-art performance and flexible scaling, it is quickly becoming a go-to tool for developers. Explore the framework and elevate your own research capabilities today.

🆔 @hackernewsgithubprojects

📦 evermind-ai/everos

Stop Losing Context: The Open Source Memory Layer for AI Agents

Tired of your AI agents forgetting everything the moment a session ends? Meet EverOS, an open source framework designed to give your agents a true long term memory. It creates a portable, self evolving memory layer that works across different AI tools and platforms. By using readable Markdown as the source of truth, it ensures your data remains accessible, versioned, and searchable. It intelligently extracts skills and cases from your actual usage, letting agents learn from past workflows without constant retraining. Take control of your agentic context and build a smarter, more persistent digital assistant today.

🆔 @hackernewsgithubprojects

📦 coleam00/excalidraw-diagram-skill

Build Smarter Diagrams with AI

Stop creating boring diagrams and start arguing visually with the Excalidraw Diagram Skill. This repository gives your coding agent the ability to generate beautiful, educational, and highly functional diagrams directly from natural language. It moves beyond simple boxes and arrows by using a render-validate loop that ensures your technical diagrams are accurate, properly laid out, and easy to read. Whether you need to map complex architectures, code snippets, or workflows, this skill uses a specialized design methodology to help you teach effectively. Install it in your agent workflow today and transform your documentation into powerful visual evidence.

🆔 @hackernewsgithubprojects

📦 mrtooher/fable-mode

Stop One-Shotting Complex Tasks: Meet Fable Mode

Are you tired of AI making basic errors on your complex projects? If your models keep missing obvious edge cases, it is time to upgrade your process. This repository introduces a Claude skill that enforces a strict, staged execution discipline. Instead of a single, risky pass, it forces the model to decompose tasks, run verification checks that can actually fail, and perform a skeptical self-critique. It acts like a rigorous checklist for your AI, ensuring tests are written and requirements are met before anything ships. Stop guessing if your work is done and start building with genuine systematic confidence.

🆔 @hackernewsgithubprojects

📦 raine/consult-llm

Get a Second Opinion From Another AI Model

Ever wonder if your AI agent is missing something obvious? When an AI reviews its own code, it often repeats the same mistakes because it shares the same underlying bias. Consult LLM solves this by letting you ping a different AI model for a truly independent second opinion right inside your existing workflow. Whether you need to debate a technical approach, review complex changes, or debug a tricky error, this tool supports multiple models and backends to provide fresh perspectives. It helps you challenge weak reasoning and find better solutions, ensuring your next coding breakthrough is just a conversation away.

📰 https://news.ycombinator.com/item?id=48525017

🆔 @hackernewsgithubprojects

📦 stonesteel27/automatiq

Stop Building Bloated Browser Bots: Use This Instead

Are you tired of heavy, fragile browser automation? Automatiq changes the game by letting you record your browser activity and turning it into a lightweight, efficient Python script. Instead of running a massive, sluggish headless browser to perform simple tasks, this tool records your interactions, uses vision analysis to understand your intent, and then has an AI agent reverse-engineer the underlying network requests for you. It runs in a native IPython sandbox, ensuring the generated code is clean and performant. Ditch the bloat, stop relying on brittle UI interactions, and start automating the web the right way.

🆔 @hackernewsgithubprojects

📦 owenyuwono/poseidon

Stunning Real-Time Ocean Simulation with WebGPU

Have you ever wanted to render a realistic, interactive ocean directly in your web browser? This project is a high-performance, GPU-driven ocean simulation built using Three.js and WebGPU. It utilizes advanced spectral ocean techniques, specifically the Tessendorf model, to achieve realistic wave behavior without the common issue of visible tiling. By running the inverse Fast Fourier Transform entirely within WebGPU compute shaders, it creates dynamic choppiness, foam accumulation, and detailed water shading in real-time. Whether you are looking to learn about complex wave synthesis or just want to explore high-end graphics on the web, dive into the code to experiment.

🆔 @hackernewsgithubprojects

📦 paca-ai/paca

Is This the Future of AI Project Management?

Are you tired of AI tools that feel like isolated chatbots? Meet Paca, a powerful, self-hosted, open-source alternative to platforms like Jira or Trello. It treats AI agents as full-fledged Scrum teammates, giving them a seat at the table to plan, execute, and adapt alongside you on the same board. By leveraging an MCP server and WebAssembly plugins, Paca allows you to fully customize workflows and integrate agents directly into your team process. It is a lightweight, free solution built to make human-AI collaboration natural and efficient. Give your AI a real job and transform your project management today.

📰 https://news.ycombinator.com/item?id=48515385

🆔 @hackernewsgithubprojects

📦 ltamann/comfyui-tbg-etur

Supercharge Your AI Images with TBG Enhanced Tiled Upscaler

Are you struggling to get ultra-high-resolution AI images without losing detail or hitting VRAM limits? This ComfyUI repository offers an advanced modular suite for tiled image generation and refinement that scales up to 100MP and beyond with precise control. It features neuro-generative tile fusion to ensure seamless, artifact-free results even during high-denoise processing. You can manage per-tile prompts and settings, use fast previews, and leverage smart caching to optimize performance. Whether you need professional-grade image polishing or creative artistic control, this toolset provides the sophisticated pipeline architecture you need to elevate your AI workflows to professional standards.

🆔 @hackernewsgithubprojects

📦 zhukunpenglinyutong/desktop-cc-gui

Stop Staring at the Terminal: Meet the Ultimate VibeCoding GUI

Are you tired of managing your AI coding workflows strictly through a cramped, text-heavy terminal? The desktop-cc-gui project changes that by providing a powerful, dedicated graphical interface for your favorite AI coding tools like Claude Code and OpenAI Codex. It solves the friction of command-line management by wrapping these engines in a clean, intuitive desktop client. You get features like session history management, real-time code diff visualization, and multi-model support all in one place. Whether you are on Windows, macOS, or Linux, it offers a seamless way to interact with AI assistants to build your projects faster.

🆔 @hackernewsgithubprojects

📦 jimmyma99/train-higgs-audio

Master Expressive AI Audio with Higgs Audio v2

Ever wondered how to make AI audio sound truly human and expressive? Meet Higgs Audio v2, a powerful foundation model that is redefining audio generation. This repository provides the essential training and generation scripts you need to fine-tune your own audio models, supporting features like LoRA training to get professional results even with limited memory. You can process custom datasets into the required format, perform zero-shot voice cloning, and even generate natural multi-speaker dialogues. Whether you are a developer or an audio enthusiast, dive into this toolkit to start building highly expressive, state-of-the-art voice applications today.

🆔 @hackernewsgithubprojects

📦 abhinand5/pi-setup

Take Control of Your AI Coding Agent Setup

Want to keep your AI coding agent environment consistent across every machine you use? This project provides a powerful way to back up and restore your agent configuration. It acts as a versioned storage layer for your extensions, themes, and custom skills, allowing you to sync your personalized setup through GitHub effortlessly. It even includes a smart file changes extension that lets you preview, accept, or revert edits before they are finalized. It is the perfect solution for developers who want a portable, synchronized workflow for their AI agent. Start syncing your environment today and reclaim your productivity.

🆔 @hackernewsgithubprojects

📦 leo-lilinxiao/codex-autoresearch

Automate Your Coding Experiments with Codex Autoresearch

Ever wish your code could improve itself while you sleep? Codex Autoresearch turns that into reality by creating an autonomous loop for your projects. You simply define a goal, like fixing bugs or optimizing performance, and the system takes over. It iterates by modifying your code, verifying the results, and deciding whether to keep or discard changes. It even handles failures by refining strategies or pivoting to new approaches entirely. Whether you are running it in the foreground or detaching it for overnight tasks, you get a fully logged experimental history. Start your next coding breakthrough today.

🆔 @hackernewsgithubprojects

📦 tile-ai/tilert

TileRT: Breaking LLM Latency Barriers

Are you tired of waiting for your large language models to respond? Meet TileRT, a revolutionary runtime engine built to push LLMs to their absolute speed limits. By decomposing complex model operators into fine-grained, tile-level tasks, this engine drastically reduces latency without sacrificing model size or quality. It is perfect for real-time applications like interactive AI and high-frequency systems that demand lightning-fast performance. With support for multi-token prediction and advanced hardware scheduling, TileRT turns heavy models into responsive powerhouses. Explore this cutting-edge approach today and experience the future of ultra-low-latency inference for yourself.

🆔 @hackernewsgithubprojects

📦 leodavinci1/kanbots

Automate Your Workflow with Multi-Agent Kanban Boards

Imagine a project board that builds itself while you sleep. Kanbots is a local collaboration interface that turns your tasks into active projects managed by AI agents like Claude Code or Codex. By assigning agents to cards, you can trigger parallel workflows, handle complex issues, or launch automated quality assurance loops. Every agent operates in its own isolated worktree, keeping your primary code safe while the AI handles drafting, testing, and implementation. With built-in features for cost management and decision-making, it transforms a standard kanban board into an autonomous team. Take control of your development cycle and let intelligent agents handle the heavy lifting.

🆔 @hackernewsgithubprojects

📦 diodiogod/tts-audio-suite

The Ultimate Voice Toolkit for ComfyUI: TTS Audio Suite

Are you tired of juggling fragmented voice tools for your creative projects? Discover the power of a unified audio pipeline with this comprehensive integration for ComfyUI. This toolkit bridges the gap by bringing together over a dozen cutting-edge AI engines, including F5-TTS, RVC, and Qwen3-TTS, into one seamless, modular workflow. You can now generate unlimited-length speech, swap character voices mid-sentence, and even use advanced audio editing tags to add natural emotion or laughter to your outputs. Stop fighting with separate software and start creating professional-grade audio productions directly in your visual workspace today for a truly streamlined creative experience.

🆔 @hackernewsgithubprojects

📦 elder-plinius/g0dm0d3

Unlock Advanced AI Performance with G0DM0D3

Tired of one-size-fits-all AI responses? Meet a powerful tool that transforms how you interact with large language models. It acts as a sophisticated API engine that races dozens of AI models in parallel to find the perfect answer for your specific needs. With built-in features like automatic parameter tuning and intelligent semantic transformation, it ensures your prompts are optimized for the best possible output. Whether you want the single best voice or a synthesized ground truth from a hive mind, this project streamlines complex workflows effortlessly. Dive in and start upgrading your AI interaction today with this liberated chat ecosystem.

🆔 @hackernewsgithubprojects

📦 khashashin/ogi

Master Your Investigations with OGI

Tired of manual link analysis slowing down your investigations? Meet OGI, an open-source visual framework designed for powerful OSINT analysis. Whether you are mapping complex connections or enriching data, OGI offers drag-and-drop visual graphs and over twenty built-in tools for tasks like geolocation, WHOIS, and social media analysis. It even features an AI Investigator that can plan your research steps and summarize findings in real time. Built for community collaboration and self-hosting, this platform helps you visualize insights faster than ever. Take control of your data and level up your investigative workflow with OGI today.

🆔 @hackernewsgithubprojects

📦 usestrix/strix

Automate Your Security Testing with Strix

Tired of manual security testing or false positives from static tools? Meet Strix, an open-source project that deploys autonomous AI agents to act like real hackers. Instead of just scanning code, Strix dynamically runs your application to find vulnerabilities and validates them with actual proof-of-concepts. It offers a full hacker toolkit, including browser automation, terminal environments, and custom exploit development. Whether you need rapid penetration testing or want to block insecure code automatically in your CI/CD pipelines, Strix delivers actionable results. Secure your applications faster and smarter by letting these collaborative AI agents handle the heavy lifting for you.

🆔 @hackernewsgithubprojects

📦 fastclaw-ai/fastclaw

FastClaw: The Ultimate Open-Source AI Agent Factory

Are you struggling to manage complex AI agents in production? Meet FastClaw, a powerful and lightweight agent factory built in Go that simplifies the entire lifecycle of your AI agents. Instead of manually wiring together containers and storage, FastClaw acts as a single-binary runtime that handles your agent's personality, memory, skills, and tools in one unified environment. It supports seamless sandbox isolation and multi-agent management, letting you deploy reliable, cloud-ready bots across various messaging platforms. Whether you are running local experiments or scaling complex workflows, FastClaw gives your agents the infrastructure they need to thrive. Start building smarter, autonomous agents today.

🆔 @hackernewsgithubprojects