📦 wangmingxuan666/tiersense

tiersense

tiersense is the decision engine that gives long-horizon AI agents a smart, self-regulating memory system. While standard agents struggle with bloated, expensive contexts, this tool analyzes conversations to provide clear control signals. It consists of three independent modules: one that scores the difficulty of the current task for smart model routing, another that identifies exactly which historical chat steps are ready to be compressed, and a final module that decides precisely when to compress or recall memories. By acting as an intelligent orchestrator, it helps developers build highly efficient agents that know exactly when to save space and when to pull up past information.

🆔 @hackernewsgithubprojects

📦 diggerhq/shipvideo

Shipvideo: Generate Launch Videos from a Prompt

Turn any website or simple text prompt into a finished launch video in under a minute. This tool skips traditional video editing entirely. Instead, a language model writes a single web page that acts like a film, using code to control every frame. A serverless agent then renders that page in a hidden browser, capturing each visual change to build the final movie file. It is a clever way to automate product demos without ever touching a timeline or camera. If you want to see how AI can replace complex editing suites with simple code, this project is a fascinating look at the future of automated content creation.

🆔 @hackernewsgithubprojects

📦 roberto02800/turnstile-token

Turnstile Token: The Missing Browser Identity

A Python tool called Turnstile Token returns a captcha token alongside the exact digital fingerprint of the browser that solved it. This matters because Cloudflare binds every token to a specific connection identity, meaning a token sent from a different client is instantly rejected as invalid. Most automation tools fail here because they only pass the string, causing the server to treat the request as if no captcha was ever solved. This library fixes that by handing back the user agent and secure network settings you need to mimic the original browser. It also automatically finds the required site key on any webpage.

🆔 @hackernewsgithubprojects

📦 shhivv/arc-cua

Arc-CUA: The Fastest Way to Automate Your Mac

Hand off boring desktop chores to a lightweight engine that clicks, types, and navigates without needing a massive, expensive A.I. model for every single move. Arc-Cua solves the slow, costly problem of traditional computer-use agents by separating big-picture planning from the actual button-pushing. It uses a smart, local decision layer to read your screen’s structure and execute precise actions, like saving a document or playing a song, in seconds. This is huge for anyone building A.I. assistants because it makes desktop automation drastically cheaper and much faster. If you are into A.I. agents, this is the kind of efficient, snappy tooling you want in your stack.

🆔 @hackernewsgithubprojects

📦 omninull/omniwm

Omni Window Manager

Omni window manager is the open-source layout organizer that brings advanced keyboard-driven tiling layouts straight to Apple Silicon Macs. It runs natively on macOS, letting you split and stack your screens without touching a mouse. What makes it incredibly cool is how it lets you switch layouts on a per-workspace basis, combining smooth side-scrolling window containers with complex spiral tree configurations. If you have ever wanted the flexible workspace customization of Linux desktop setups on your Mac, this application makes it incredibly easy to manage a multi-monitor layout. Go grab the latest build and reclaim your screen space.

🆔 @hackernewsgithubprojects

📦 siliconlabai/openjev

OpenJev: The Decision Playground

OpenJev is a browser playground that turns messy situations into structured, confident answers. The most surprising part is that it can run on a single graphics card using a specific reinforcement learning model called Decider, rather than relying on massive cloud APIs. It works by taking a simple text description and a set of questions, then using three different backends to score every option. This means you can test how different models interpret your data in real-time. It is perfect for developers who want to see exactly how an AI weighs choices without writing complex integration code. It gives you a clear, visual look at the decision-making process.

🆔 @hackernewsgithubprojects

📦 446599/ccodex-rotate

ccodex-rotate: The Local Proxy Chaser

Ccodex-rotate is a local tool that quietly hunts for valid access windows so your AI requests never drop to a weaker model. It runs on your Mac or Windows machine, sitting between your app and the internet without touching your system settings. The real magic is its background hunter, which constantly scans different network paths every couple of minutes to find the exact route serving the full-power model. It then automatically swaps your traffic to that path and injects fresh, short-lived credentials to keep you on the top tier. It is not a permanent fix, but a constant chase that keeps your session running at full speed.

🆔 @hackernewsgithubprojects

📦 wquguru/dasheng

Dasheng: The Local Tool That Scores Your English Reading

Dasheng is the self-hosted reading tool that tells you exactly which words you got wrong. Most pronunciation apps send your voice to a cloud server, but this runs entirely on your own computer. It listens to you read a passage and uses a local model to spot mistakes in real time. As you speak, the correct words light up green, while any errors get marked in red immediately. This gives you instant feedback without any monthly fees or data leaks. It is a great way to practice English privately and see your progress clearly.

🆔 @hackernewsgithubprojects

📦 shayanabbas1/dbdelve

dbdelve

dbdelve is the modern database client that finally brings instant, fluid scrolling and a tiny memory footprint to your SQL workflows. Built entirely in Rust using the high-performance GPUI framework, it runs circles around bloated, resource-heavy electron apps. It natively supports Postgres, MySQL, SQLite, Snowflake, and SQL Server with lightning-fast navigation across millions of rows. What makes it truly special is its strict, built-in safety gates that prevent you from accidentally running destructive queries, alongside separate environments that keep your staging and production tabs completely isolated. It is keyboard-first, runs entirely local, and puts you in complete control of your data.

🆔 @hackernewsgithubprojects

📦 mak5er/aircard

AirCard

Customize your iPhone lockscreen and digital wallet without needing a jailbreak. Aircard lets you completely change the look of your Apple Pay cards and lockscreen passcode buttons directly from your Mac using a built-in exploit. You can assign custom artwork and logos to individual payment cards or apply uniform skins in a single click. Even cooler, you can build custom lockscreen passcode themes by slicing up a single wallpaper across the keypad grid or designing individual keys. It is fully self-contained, works with standard theme files, and lets you personalize your device safely with absolute ease.

📰 https://news.ycombinator.com/item?id=49864670

🆔 @hackernewsgithubprojects

📦 gulelmatthews/polymarket-perpetual-bot

Polymarket Prediction Bot

This terminal dashboard hunts for free money inside prediction market spreads. Polymarket Prediction Bot watches live order books to spot when a yes or no price is wrong, flagging arbitrage opportunities before you even blink. It simulates trades so you can test your edge without risking a single dollar, keeping your wallet safe while you learn the ropes. Think of it as a probability scanner for your terminal, built for anyone curious about how real money moves on prediction markets. It is an unofficial community project, not financial advice, but it is a fascinating window into automated trading logic.

🆔 @hackernewsgithubprojects

📦 deepakness/cogsend

CogSend: Own Your Social Media Schedule

CogSend solves the frustrating problem of constantly rewriting your social posts for different platforms. Instead of retyping your content for Mastodon, Bluesky, or LinkedIn, you write your update just once in a shared draft editor. The software automatically splits longer text into threads that fit each network’s specific rules, so you never have to guess where to break lines. Since the entire application runs on your own cloud account, your data stays private and you avoid monthly subscription fees. It is the perfect tool for creators who want control without the hassle. Stop renting your audience and start publishing with complete ownership.

🆔 @hackernewsgithubprojects

📦 liyucheng1997/332_lab-jev-chat

Jev Chat Assistant

A floating window on your phone reads your live WeChat conversation and tells you exactly what your friend is really thinking. This project, called Jev Chat Assistant, acts like a digital co-pilot that analyzes the text on your screen to guess their true intent. It then uses an AI model to draft three possible replies for you to pick from. The coolest part is that it never sends anything automatically. It only fills the input box, so you stay in total control of the final send button.

🆔 @hackernewsgithubprojects

📦 stew675/llama-cpp-rdna-boosts

Squeeze More Speed From Your AMD GPU

Run local AI models on your AMD graphics card at blazing speeds without sacrificing output quality. This project delivers a focused set of performance patches for llama dot cpp, the popular tool for running language models locally. The core value is high-speed inference on RDNA architecture GPUs, like the RX 7000 and 9000 series, while maintaining numerical assurance. It solves the problem of slow or unstable performance on these specific chips by optimizing the underlying compute kernels. You can apply the whole set or pick specific blocks to boost your hardware. It is a practical toolkit for anyone wanting to get the most out of their existing hardware.

🆔 @hackernewsgithubprojects

📦 togethercomputer/tev1

Tev1: The $17 Decision Model

Tev1 is the open-weight decision model that turns a tiny language model into a precise classifier for a fraction of the cost. It takes a specific situation, a question, and a short list of choices, then returns exactly one letter answer without any extra chatter. This project is built on a smaller base model and trained to handle real world judgments like checking return policies or sorting support tickets. The repository gives you the full recipe, training data, and saved results so you can replicate the exact workflow. You can build and deploy your own custom decision model in just a few steps.

🆔 @hackernewsgithubprojects

📦 n1m21n/infinite

Infinite: The Node-Based Audiovisual Workstation

Infinite is the node-based workstation that finally puts video, audio, and 3D graphics on the same canvas. It solves the messy problem of needing five different apps to make one visual track, letting you patch everything together in real time with your GPU. The coolest part is the predictive engine, which actually listens to your performance and learns your style to suggest the next note or color shift before you even think of it. It is basically a live coding playground for musicians who want total control.

🆔 @hackernewsgithubprojects

📦 barty-bart/motion-graphics

Motion Graphics

Motion Graphics is the automation tool that turns boring talking-head videos into beautiful, custom B-roll timed perfectly to your words. By connecting directly with AI code assistants, this utility analyzes your video layout and transcript to design, animate, and render sleek, continuous overlays. Instead of jump cuts, it uses a single morphing shape driven by an on-screen cursor that shifts from buttons to charts in one elegant motion. Best of all, it handles the tedious setup by exporting ready-to-use video files and transparent overlays that you can slide straight into your favorite editor.

🆔 @hackernewsgithubprojects

📦 x-square-robot/xplanner

X-Planner: The Robot's Mental Map

X-Planner gives physical robots a structured way to think about complex tasks, which is a big deal for making machines actually useful in the real world. Instead of just reacting to what they see, this system breaks down big jobs into clear, step-by-step plans. It learns from human demonstrations to understand what success looks like and how to recover when things go wrong. This makes robot behavior much more predictable and easier to trust. If you are into robotics, this is the kind of thinking that turns clumsy machines into capable helpers, so it is definitely worth a look.

🆔 @hackernewsgithubprojects

📦 tvollstaedt/exodium

Exodium: Your Retro Game Library

Exodium turns your computer into a personal museum of retro gaming. Instead of forcing you to download a half-terabyte archive just to look at one game, it lets you stream specific titles directly from the community torrents. You get to browse thousands of pre-configured DOS, Windows, and adventure games, grabbing only what you want to play right now. It even handles the emulator setup for you, so you can skip the technical headache and just press play. This is the perfect way to rediscover gaming history without filling up your hard drive. Grab the latest release and start exploring the collection today.

🆔 @hackernewsgithubprojects

📦 agricidaniel/jev-seo

Jev SEO: AI-Powered Website Audits

Get a complete, live website health check in under four minutes. Jev SEO crawls your site to run fifty-two specific technical checks while an AI model reads every page to judge real content quality. It combines these findings into a prioritized list of fixes, then generates a professional PDF report and an Excel action tracker. This tool is useful because it moves beyond simple link checking to evaluate whether your pages actually help users. It is interesting to see how the AI assesses search intent and page trustworthiness. You get clear, actionable steps rather than vague advice.

🆔 @hackernewsgithubprojects