📦 royerlab/kaggle-cell-tracking-competition

Kaggle Cell Tracking: How We Follow Cells in 3D Video

Kaggle Cell Tracking Competition is the neural network baseline that solves the headache of tracking cells in 3D time-lapse microscopy where the answers are missing. This project takes raw 3D video frames and uses a specialized deep learning model to first spot every cell and then link them across time, even when the ground truth data is sparse and incomplete. It is particularly clever because it ignores the parts of the video without annotations during training, focusing only on what is known to build a robust model. This is useful for anyone in bioinformatics who needs to automatically track cell division and movement without manually labeling every single frame.

🆔 @hackernewsgithubprojects

📦 4evy/pared

Pared: Delete Apple Intelligence Models Without Disabling SIP

Pared is the command-line tool that removes Apple Intelligence models without disabling system protections. Most deletion scripts require turning off SIP or booting into recovery mode, which is risky and annoying. Pared works while your security stays fully on. It finds the large AI files Apple downloads to your Mac and asks Apple’s own background service to delete them properly. Because it uses Apple’s official method, your system remains safe and stable. You can also manage which features stay active using a simple text file, making it easy to automate. This tool gives you back your hard drive space while keeping your Mac secure.

🆔 @hackernewsgithubprojects

📦 llopresto87/cypress

Cypress: The AI Coding Team In A Box

Cypress is the multi-agent seed that gives your AI coding tools a real expert team instead of just a chatbot. It drops into any project to create a roster of specialized agents, from architects to security testers, each with their own strict rules. The coolest part is the progressive discovery graph, which lets these agents find exactly the knowledge they need without getting lost. It works with any coding assistant you use, so you get disciplined, spec-driven coding without locking yourself into one vendor. If you want your AI to work like a careful team of professionals rather than a guesser, this is the setup you want.

🆔 @hackernewsgithubprojects

📦 isene/scribe

Scribe: The Terminal Editor That Thinks Like a Writer

Scribe is a terminal text editor that boots in under ten milliseconds and treats your words like precious cargo, not just data. It strips out the complex programming tools you don’t need and replaces them with a built-in reading mode that dims the surrounding text so you can focus purely on the sentence in front of you. This tool also lets you hand specific paragraphs directly to an AI assistant for instant rewrites, all without ever leaving your terminal. It is a single, static binary that starts instantly and respects the speed of your thought process.

🆔 @hackernewsgithubprojects

📦 phalladar/macdirstat

MacDirStat: See What's Eating Your Mac's Storage

Find out exactly where your Mac's storage went by turning your hard drive into a visual map. MacDirStat scans your files and draws them as a single, colorful image where the size of each rectangle matches the space that file uses. It's like a puzzle that shows you your whole computer at once, making it easy to spot the giant videos or old backups hiding in plain sight. Because it runs entirely on your device without any internet connection, it keeps your data private while helping you clear out clutter. Just look for the biggest blocks to find the heavy hitters.

🆔 @hackernewsgithubprojects

📦 mcg-nju/onestreamer

OneStreamer: The Video AI That Knows When to Speak

OneStreamer is the streaming video model that finally answers questions before you even ask them. Most video AI just watches, but OneStreamer actively decides when it has seen enough to respond, like counting objects as they appear. It solves the problem of waiting for the end of a video to get an answer, giving you real-time, proactive updates. This is wild because it means the system is truly interacting with the stream, not just processing it. You get instant answers without the delay, which changes how we think about video understanding. It’s a clever way to make AI feel more responsive and actually useful in the moment.

🆔 @hackernewsgithubprojects

📦 leonxlnx/cinetic

Cinetic: Your Code Agent Directs a Movie

Your coding agent can now direct a polished launch film from a simple text prompt. Cinetic stops it from picking the same boring animations by randomly drawing from a massive library of cinematic moves. It builds the entire video, including the music and sound effects, right from your codebase. The result is a unique, professional-looking ad that feels handmade, not generated. It is the closest you will get to having a movie studio in your terminal.

🆔 @hackernewsgithubprojects

📦 jeantimex/tokyo

Procedural Tokyo

Procedural Tokyo is the browser project that builds a living city from public data. It takes real maps and elevation surveys to place every building and road accurately, then adds generated details like traffic, birds, and shifting daylight. You can fly through neighborhoods like Shibuya or watch trains run their actual routes in real time. It turns a flat map into a dynamic world you can explore right in your browser without installing anything. If you want to see Tokyo move and breathe on your screen, this is the place to go.

🆔 @hackernewsgithubprojects

📦 kushneryk/join.cloud

Join Cloud: Slack for AI Agents

AI agents now have their own Slack, a shared workspace where they message each other and commit files to shared storage in real-time rooms. This project, Join Cloud, solves the problem of isolated agents by giving them a collaborative hub where they can coordinate tasks and even review each other's work. It supports standard protocols like MCP and A2A, meaning you can connect your favorite agents without custom code. Whether you are building a multi-agent system or just want to see agents cooperate, this gives you a simple way to watch them work together live.

🆔 @hackernewsgithubprojects

📦 mariagorskikh/open-instinct

Open Instinct: The AI Agent You Text

Text your own computer like it is a friend, and watch it book dinner, write files, and pay bills without ever touching your credit card. Open Instinct is an open source personal agent that lives in a tiny cloud machine, reachable only through an iMessage thread. The coolest part is the trust system: you decide exactly what your partner, friends, or strangers can ask it to do, so your calendar stays private while your agent still helps you coordinate dinner. It is not a chatbot; it is a worker with a desktop, a wallet, and strict rules.

📰 https://news.ycombinator.com/item?id=49966816

🆔 @hackernewsgithubprojects

📦 therunaway5/phase-distorter

Play EarthBound on Your Monitor

Take the cult classic EarthBound and play it natively on your own computer without needing a retro console or an emulator. Phase Distorter is a C plus plus port that brings the SNES hit to Windows and Linux, letting you run the game directly on your desktop. The coolest part is how it handles the visuals. It stretches the original four by three aspect ratio into widescreen, ultrawide, or even super ultrawide formats. It does this without breaking the gameplay, actually rendering new scenery to fill those extra pixels. It also includes an optional C R T filter so it feels authentic.

🆔 @hackernewsgithubprojects

📦 yozorakurage/yolupainter

YoluPainter Paints 3D Models

YoluPainter is a drawing app that lets you paint textures directly onto three-dimensional models. The coolest part is a live link to Unity. You load your character model once, and every stroke you make shows up instantly on the avatar in the scene. You are painting the actual material, not a flat image. This means you can see exactly how your colors look on the final character as you work. It solves the annoying problem of guessing how a texture will wrap around a shape. If you make games or characters, this cuts the back and forth between software. It is a free, open-source tool built in Rust.

🆔 @hackernewsgithubprojects

📦 toorux/steam-frame-6ghz-tool

Steam Frame 6 GHz Tool

Steam Frame 6 GHz Tool lets you force a USB Wi-Fi adapter to run in US regulatory mode without touching any driver files, which is handy because standard settings often block the newer six gigahertz band. It is a simple desktop app that finds the adapter, flips the region switch, and even remembers the setting across reboots if you want. It also has a neat feature to talk directly to your headset over a secure connection to fix wireless settings there, so you are not stuck guessing. If you own this hardware and want full bandwidth, this is the clean way to do it.

🆔 @hackernewsgithubprojects

📦 camenduru/tostai-sprite-sheet-studio

TostAI Sprite Sheet Studio: Video to Sprite

TostAI Sprite Sheet Studio is the web tool that turns raw video clips into game-ready sprite sheets. Most developers have to manually cut frames and clean up backgrounds, but this app automates the entire messy process. It uses a specialized background removal model to isolate your character, then snaps the pixels to a grid so they look crisp and professional. This is ideal for game developers who want pixel-perfect animation without spending hours in image editors. If you are building a game and need consistent sprites from recorded motion, this is exactly what you need to speed up your workflow.

🆔 @hackernewsgithubprojects

📦 cobanov/ps5lm

PS5LM: Run LLMs on a PS5

PS5LM is the jailbreak tool that lets you run full language models directly inside your PlayStation 5. It ports the popular inference engine onto the console so you can download, run, and chat with open-source models using your game controller. The project focuses on making these models accessible without needing a computer, turning your living room console into a local AI assistant. While it currently supports smaller models to stay within system limits, it proves that powerful artificial intelligence can live entirely on your hardware.

🆔 @hackernewsgithubprojects

📦 poisonxa16/pxa

Run AI on Old GPUs with PXA

Run large language models on older NVIDIA graphics cards that most modern software ignores. PXA is a local engine built specifically for chips like the Tesla P100 and GTX 1080 Ti, which are no longer targeted by new tools. It squeezes more speed and quality from these aging GPUs by using custom math routines and smarter data packing. The standout feature is its auto-tuning: it measures your specific hardware and picks the fastest settings automatically, so you do not need to be an expert. You get a simple web interface to start a chat server in seconds, no coding required.

🆔 @hackernewsgithubprojects

📦 naklitechie/localmind

LocalMind

LocalMind runs a private AI assistant entirely inside your browser tab, with no server and no account. It handles chat, images, voice, and documents using models that execute directly on your device. This solves the privacy problem by keeping your data local. It is interesting because it gives you a powerful, offline AI companion without uploading your files to a cloud. You get full control and speed. The tool is self-contained and secure. It is a clever way to own your AI experience. Keep this in mind. The best AI is the one that never leaves your house. This is the future of personal tech.

🆔 @hackernewsgithubprojects

📦 adamnizol/comfyui-anima-enhancer

ComfyUI Anima Enhancer

A single custom node can boost your image generation speed by thirty-five percent while sharpening fine details. ComfyUI Anima Enhancer is a plugin that improves outputs from Anima models by replaying specific internal blocks during the denoising process. It solves the common issue of blurry or incoherent results by focusing the model’s attention on key structural layers. You can also enable Spectrum mode for a significant speed boost without sacrificing quality. Just plug it into your workflow before the sampler and adjust the block indices to taste. If you use Anima models, this is a quick, effective upgrade worth trying.

🆔 @hackernewsgithubprojects

📦 framefilter/keyroost

Keyroost: One App for All Security Keys

Keyroost lets you manage a YubiKey, a SoloKey, or a Token2 fob with one single app, no matter who made it. Most tools only work with one brand, but this works because it speaks the universal language of open standards instead of relying on private software from each maker. It handles everything from setting up fingerprints to generating encryption keys and resetting a stuck device. The coolest part is the factory reset: it wipes every part of the key in the right order, so you can start fresh without guessing.

🆔 @hackernewsgithubprojects

📦 benoitc/erlang_wasm

erlang_wasm: The WebAssembly Runtime Written in Erlang

erlang_wasm is the WebAssembly runtime that runs entirely inside the Erlang virtual machine without any heavy native dependencies. It lets you execute code compiled to WebAssembly directly inside your Erlang applications, which solves the tricky problem of safely running untrusted or multi-language code without leaving your existing system. The really surprising part is how it handles memory, using clever techniques to make it fast enough to compete with runtimes written in C. If you work with Erlang and need to run external code, this gives you a powerful, contained way to do it.

🆔 @hackernewsgithubprojects

📦 magicbug/oscarwatch-tracker

OscarWatch Tracker: Satellite Pass Automation for Ham Radios

OscarWatch Tracker automatically controls your radio frequency and antenna rotator to track moving satellites in real time. This desktop application solves the problem of manually calculating complex Doppler shifts and aiming antennas for brief satellite passes. It calculates the exact frequencies needed for uplink and downlink signals while accounting for the satellite's speed and your specific location. For amateur radio operators, this means less mental math and more focus on making clear contacts. The software supports various radio models and rotator controllers, handling the technical tracking work in the background. You get a reliable way to track these fast-moving objects without constant manual adjustments.

🆔 @hackernewsgithubprojects