📦 66hex/frame

Frame

Stop wrestling with complex command line arguments when you need to convert your media. This sleek tool gives you a beautiful, native desktop interface to run incredibly detailed FFmpeg conversions without writing a single line of terminal code. You can easily drag in your videos, audios, or photos, and get total, granular control over everything from codec settings and audio filters to burning in subtitles. The absolute best part is the native preview panel, which lets you visually crop, trim, and drag an image overlay right onto your video before you convert. It is an incredibly fast, lightweight utility that makes professional-grade media processing accessible to anyone who wants to...

🆔 @hackernewsgithubprojects

📦 andrewrabert/jellium-desktop

Jellium Desktop

Control your home media server with native performance by running a dedicated player that blends web and local code. jellium-desktop builds an unofficial desktop app for Jellyfin using Rust, embedding Chromium to render the interface while routing the actual video rendering through a highly optimized media player. By layering a web-based menu as a transparent texture directly over native video hardware, it delivers fluid playback without the stutter and overhead of a standard web browser. It is a brilliant way to enjoy your movies and shows with complete media-key support and smooth performance across Linux, Mac, and Windows.

🆔 @hackernewsgithubprojects

📦 github/gh-stack

gh-stack

You can now break your massive, intimidating pull requests into a neat chain of tiny, bite-sized branches that build on top of each other. The gh-stack project is a clever official GitHub CLI extension that automates the absolute worst parts of managing stacked pull requests. Instead of manually rebasing five different branches, updating their base targets on GitHub, and losing your mind to merge conflicts, this tool handles the entire pipeline. You can visualize your stack, safely cascade rebases from the bottom up, and even use a full terminal interface to visually reorder, rename, or merge branches with simple keystrokes. It makes code reviews fast and painless again.

🆔 @hackernewsgithubprojects

📦 eyadkelleh/awesome-claude-skills-security

awesome-claude-skills-security

awesome-claude-skills-security is the security toolkit that brings massive cybersecurity databases directly into your everyday coding assistants. Instead of manually downloading gigabytes of security data, this project packages curated lists of common passwords, fuzzing payloads, and web shells as smart skills for tools like Cursor and Claude Code. You can literally ask your AI assistant to run an authorized SQL injection test or scan your workspace for exposed API keys using simple, everyday language. It even includes specialized prompts for testing AI safety and bias. It is a fantastic bridge between automated coding and professional security auditing.

🆔 @hackernewsgithubprojects

📦 opusgamelabs/game-creator

Game Creator

You can now transform a single tweet or simple text prompt into a fully playable, monetized browser game in just ten minutes. A clever tool called game-creator makes this possible by turning AI coding assistants into an autonomous game studio. It scaffolds your game using Phaser for 2D or ThreeJS for 3D, automatically designs the visual polish, generates procedural retro audio, and instantly deploys it to the web. The best part is the built-in QA agent that tests the gameplay in a headless browser at every step, automatically patching bugs on the fly. It is a brilliant way to instantly bring your weirdest game ideas to life.

🆔 @hackernewsgithubprojects

📦 hope7happiness/t2i-release

t2i-release

Generate gorgeous images directly from text without dealing with overly complex setups. The t2i-release repository is a clean PyTorch re-implementation of MiniT2I, a minimalist text-to-image generator that works entirely in pixel space. Instead of relying on complicated image tokenizers, multi-stage pipelines, or extra training losses, it strips the process down to a straightforward denoiser conditioned on a frozen text encoder. It is highly practical because it gives you everything you need to run quick inferences, train with custom datasets, or adapt the model with adapters. You get a lightweight, easy-to-run system that proves simple recipes can still deliver incredible results.

🆔 @hackernewsgithubprojects

📦 quarqlabs/agent-oss

Argus Agent

Most AI agents have a massive problem with memory, either forgetting what you told them or mixing up when things actually happened. But agent-oss, also known as Argus Agent, completely fixes this with a local-first cognitive runtime built specifically for bulletproof long-term memory. Instead of just doing basic vector search, it acts like a disciplined brain, organizing what it learns into semantic facts, episodic events, and procedural behavioral rules. It even runs self-correcting retrieval passes if it realizes it is missing key information. It is essentially the ultimate, private memory companion that actually knows the difference between when a memory was saved and when it actually happened.

🆔 @hackernewsgithubprojects

📦 raketenkater/llm-server

llm-server

Running massive open-source AI models across a mismatched mix of graphics cards usually ends in a frustrating mess of system crashes, out-of-memory errors, and tedious manual configuration. The llm-server project solves this by acting as an intelligent, auto-tuning launcher that acts like a smart conductor for your hardware. It reads your specific computer setup, analyzes your graphics cards and system memory, and builds a precise placement plan to split large models safely across your hardware. It even measures safe performance tweaks behind the scenes to find the fastest, most stable settings for your exact machine, giving you smooth local AI without the headaches.

🆔 @hackernewsgithubprojects

📦 rohithgilla12/data-peek

Data Peek

Data-peek is the lightweight database client that finally rescues you from bloated, slow-loading development tools. Instead of waiting ages for heavy enterprise software to start up just so you can check a single row, this desktop app launches in under two seconds to let you quickly view and query your data. It supports PostgreSQL, MySQL, SQL Server, and SQLite, and it includes a brilliant watch mode that auto-refreshes your queries with real-time, color-coded diff highlights as the data changes. It is the ultimate tool for developers who want to get in, get their answers, and get right back to coding.

🆔 @hackernewsgithubprojects

📦 ai-builders-foundation/ai-builders-curriculum

AI Builders Curriculum

The educational toolkit that finally makes building AI apps approachable, ai-builders-curriculum is a free, vendor-neutral guide for crafting full-stack software from scratch. Instead of locking you into expensive platforms, it teaches you the fundamentals of databases, user accounts, and secure APIs before plugging in the smart stuff. The best part is the offline starter kits, which let you run a complete document-chatting app or a smart note summarizer locally on your computer with absolutely zero API keys. It is the ultimate hands-on roadmap to go from an empty folder to a fully deployed application over a single weekend.

🆔 @hackernewsgithubprojects

📦 mohitsoni48/turbo-llm

Turbo LLM

Turbo-llm is the lightweight local AI manager that finally lets you swap out your underlying generation engine on the fly. Instead of locking you into one stock runtime, this offline-first tool lets you point to any community fork or custom binary to run your models. It automatically benchmarks your GPU on load to find the fastest settings, giving you real, measured speed boosts of up to double the performance on the exact same hardware. With its polished web interface and built-in compatibility for both OpenAI and Anthropic APIs, you can point advanced coding agents at your machine with a single command.

🆔 @hackernewsgithubprojects

📦 dl1683/ant-irys

ant-irys

An open source project named ant-irys can beat elite, fine-tuned AI models on complex professional benchmarks for just one dollar and thirty cents per task. While expensive frontier systems fail on complex multi-step reasoning, this framework coordinates tiny, inexpensive models to work together like an analytical team. Instead of forgetting everything between turns, it builds a persistent, structured blackboard that tracks every document, observation, and logical gap over time. This architectural approach proves that structured, stateful coordination is far more important than raw model size. You can run deep, auditable document analysis on a budget, shifting AI from an expensive single-shot guess to a lasting knowledge base.

🆔 @hackernewsgithubprojects

📦 yuliang-liu/monkeyocrv2

Monkey OCR v2

The visual-text foundation model that finally makes document AI feel effortless across seventeen different languages, monkeyocrv2 is a major step forward for anyone trying to pull real meaning out of messy layouts. Instead of struggling with separate tools for finding text, reading math formulas, and detecting tampered documents, this project unites fine-grained text modeling and cross-task learning into a single, highly efficient vision encoder. It is incredibly useful because it tackles real-world chaos, from overlapping text to complex multilingual layouts, without breaking a sweat. Give it a spin to see how easily it translates documents into structured, usable data.

🆔 @hackernewsgithubprojects

📦 sylearn/aiusage

AIUsage

Manage all your separate AI subscriptions from a single macOS dashboard. This clever native app lets you track quotas and costs across over ten different providers while hot-swapping your active API nodes instantly. Instead of restarting your developer tools or editing messy configuration files by hand, you can route command-line tools like Claude Code, Codex, and OpenCode through customizable local proxies. It even lets you use these coding assistants with any OpenAI-compatible models, giving you complete freedom to run alternative backends. It is the perfect way to monitor your developer credits and keep your local workflows organized.

🆔 @hackernewsgithubprojects

📦 artymclabin/gmail-mcp-server

Gmail MCP Server

The gmail-mcp-server is the smart tool that lets you control your entire inbox directly through local artificial intelligence assistants. Instead of jumping between tabs or pasting text back and forth, this clever background helper lets you search, read, and organize your emails using simple conversational language. The absolute best part is how it handles attachment downloads and full message exports to your computer without clogging up the AI with heavy background files. You can draft messages, manage labels, and even run batch operations safely from your terminal, giving you complete hands-free control of your daily communications.

🆔 @hackernewsgithubprojects

📦 jimliu/baocut

BaoCut

BaoCut is the open-source agent skill that finally lets your AI coding assistants drive a full video editing and subtitle pipeline directly from the command line. Instead of clicking through menus, you can just tell your developer agent to transcribe a video, translate subtitles into Chinese, or automatically clean up silence and filler words from a talking-head clip. It seamlessly hooks into tools like Claude Code and Codex, translating your natural language requests into precise local commands that run right inside the native macOS app. It is like giving your favorite AI assistant a professional video editor's toolkit to polish, translate, and export clean media in seconds after you ask.

🆔 @hackernewsgithubprojects

📦 maxprokopp/film-space

Film Space

Record cinema-grade camera movements right from your living room by walking around with your phone. A new project called film-space turns your device into a virtual camera rig. You place virtual human stand-ins to block out your scene on a digital grid, then switch to camera mode. As you physically walk, tilt, and turn, the app uses augmented reality to translate your real-world steps into virtual camera paths. By recording these custom camera moves, you get the perfect reference video to feed into generative AI tools, giving you complete control over your final cinematic shots.

🆔 @hackernewsgithubprojects

📦 thatmagicalcat/txm

txm

txm is the terminal math rendering engine that finally brings beautifully formatted LaTeX equations straight into your command line. Built in Rust with convenient Python bindings and Neovim integration, this clever tool parses complex math syntax and renders it perfectly using unicode and terminal graphics. Instead of squinting at raw, unreadable LaTeX markup in your terminal, you can run a single command to see fractions, square roots, and integrals instantly displayed in high fidelity. It is a game-changer for terminal-dwelling developers and researchers who want to quickly verify formulas without leaving their shell. Give it a run to make terminal math look gorgeous.

🆔 @hackernewsgithubprojects

📦 xullexer/pydns-scanner

pydns-scanner

The pydns-scanner is the high-performance DNS scanner that finally packages massive IP scanning inside a gorgeous terminal user interface. Instead of staring at boring command-line walls, you get a beautiful, interactive dashboard built with Textual to hunt down working DNS servers across millions of addresses. Its cleverest trick is a bidirectional multi-range scan strategy, meaning it tackles IP blocks from the front and back simultaneously with a random shuffle to find active servers incredibly fast. It even auto-scores DNS types and handles proxy testing without breaking a sweat. It is the ultimate diagnostic companion for your terminal.

🆔 @hackernewsgithubprojects

📦 braedonsaunders/homerun

Homerun

Homerun is the open-source prediction market platform that finally gives you full-code Python control over your trading strategies. Instead of forcing you into rigid templates or toy rules, this platform lets you write custom async Python classes to scan, backtest, and trade live opportunities across Polymarket and Kalshi. It features a brilliant, microstructure-aware simulator that uses actual order book history and real-world latency to mimic true execution before you risk a single dollar. You can design custom data sources, run advanced backtests, and seamlessly transition from shadow testing to live trading using a clean, unified dashboard. It is the ultimate playground for building prediction market alpha.

🆔 @hackernewsgithubprojects

📦 radiant303/springnote

SpringNote

SpringNote treats your daily work notes as a living, growing system rather than just static text files. Built with Flutter and Rust, this local-first desktop app connects to models like DeepSeek to automatically transform your scattered daily thoughts into structured weekly and monthly reports. It even features a chat interface called the Memory Book, allowing you to search and talk directly to your past records like a personal knowledge base. There is even a built-in hourly wage clock that calculates your earnings in real time while you work. It is the ultimate tool to turn daily grinding into organized, reusable knowledge.

🆔 @hackernewsgithubprojects