📦 addxai/reflectworld

ReflectWorld: Giving AI a Real Memory

ReflectWorld gives artificial intelligence agents a persistent visual memory that actually works. It connects to live camera feeds, like those on your phone or a robot, and stops treating every moment as a brand new, isolated clip. Instead, it builds a continuous, searchable map of your real world, tracking specific people, objects, and events over time so the agent knows who is who and what changed. This is huge because it moves AI beyond just describing a single scene to truly understanding the ongoing story of a place. Your data stays on your machine for privacy, and the system learns from your actual footage.

📰 https://news.ycombinator.com/item?id=49195953

🆔 @hackernewsgithubprojects

📦 spatiaos/procedura

Procedura: Text to Editable 3D Code

Generate 3D models from a simple text prompt that you can actually edit like code. Procedura solves the frustrating problem where AI usually spits out a frozen block of geometry that you cannot change without starting over. This tool uses an AI agent to write a parametric program, meaning every part is a named variable you can adjust, scale, or move later. It also adds realistic materials and moving joints, so your model isn’t just a static statue. The coolest part is that it’s not a black box; you get the actual source file to tweak.

🆔 @hackernewsgithubprojects

📦 dmtrkovalenko/bashka

Bashka: The Safety Guard for Bash Scripts

Bashka is the safety guard that finally stops you from blindly piping unknown scripts into your terminal. Instead of running an install command immediately, it reads the code first to check for dangerous actions like stealing passwords or hiding malicious files. If it finds something scary, it blocks the install and explains exactly what went wrong. It also tracks every program you install this way, making cleanup easy later. It’s a simple drop-in tool that turns a risky habit into a secure, transparent process, giving you total control over your system.

📰 https://news.ycombinator.com/item?id=49722348

🆔 @hackernewsgithubprojects

📦 dabit3/jev-experiments

Jev Experiments: Sub-Second AI Code Checks

Jev Experiments is the collection of demos that proves AI code checks can actually finish before you finish your sentence. The standout piece is Commit Sentry, which scans your staged code changes in under half a second to catch real disasters like leaked passwords or deleted database tables. Traditional linters miss these semantic traps, while standard AI reviewers take too long to be useful before you commit. This tool sits in the middle, analyzing your changes in parallel to block dangerous code instantly. It turns a slow, after-the-fact review into a fast, real-time safety net.

📰 https://news.ycombinator.com/item?id=49757995

🆔 @hackernewsgithubprojects

📦 ekzhang/openjev-sglang

Openjev: The One-Token Classifier

Openjev is the self-hosted inference server that turns a massive language model into a lightning-fast classification engine. It runs a specific Qwen model on high-end graphics hardware but completely skips the slow, token-by-token generation process that makes chatbots so sluggish. Instead, it reads the model's internal confidence signals from the very first token, letting it answer dozens of yes, no, or multiple choice questions in a single, near-instant burst. This turns a general-purpose AI into a precise, low-cost tool for tasks like routing customer support tickets or grading exam answers without paying for expensive API calls.

🆔 @hackernewsgithubprojects

📦 fathom-media/fathom

Fathom: The All-In-One Media Client

Fathom turns your desktop into a single, powerful media center for your entire Jellyfin server. You can manage users, libraries, and live TV directly from the app, so you rarely need to touch the web dashboard. The real surprise is the built-in YouTube player that runs right alongside your personal movies and shows. It uses the exact same high-quality controls and hardware decoding for both, giving you one seamless experience without ads or account requirements. It’s like having a premium TV guide and a private streaming service in one neat package. If you want to stop juggling multiple apps, this is your new home base.

📰 https://news.ycombinator.com/item?id=49758543

🆔 @hackernewsgithubprojects

📦 xgen-labs/xgen-jing

Jing: First-Person AI Worlds

Jing is the interactive model that lets you walk through AI-generated worlds as if you were actually there. It turns simple text and images into first-person video and audio, so you can move forward, turn, and talk to characters in real-time. This is not just a static video; it is a living experience where your actions change what you see and hear. It solves the problem of passive media by creating a playable, dynamic environment from a simple story prompt. The result is a fascinating glimpse into the future of immersive, personalized digital experiences.

🆔 @hackernewsgithubprojects

📦 lunarresearcher/copy

COPY: The Terminal That Hacks Robinhood Chain

A single terminal command turns a chaotic crypto market into a live, automated trading desk. This tool, called Copy, monitors Robinhood Chain to find new tokens and profitable traders. It instantly scores every setup against strict rules, telling you exactly why a trade passes or fails. The most surprising part is that it can automatically open and close fake positions to test your strategy without risking a dime. If you want to watch smart money in real time without the stress, this is your new command line.

🆔 @hackernewsgithubprojects

📦 avivsinai/agent-message-queue

agent-message-queue

Agent Message Queue turns your computer's file system into a live, crash-safe messaging hub for AI coding partners. No database or server required, just plain text files that never corrupt, even if your machine crashes mid-send. It solves the chaos of multiple AI agents working on the same codebase by giving them a reliable, human-readable channel to coordinate tasks and review each other's work without needing a human referee. You can inspect every message with a simple text editor or version control tool, making the whole system transparent and debuggable.

🆔 @hackernewsgithubprojects

📦 unclebob/uml-viewer

UML Viewer

UML Viewer turns your code into a live map that instantly exposes which parts of your app are fragile. It scans your source files to draw a diagram where every box is colored by real health metrics, so you see exactly which areas need attention before you even open the code. It highlights dependencies that break your architecture rules in bright red, giving you a clear visual signal of where the design is straining. This is not just a static picture, but a dynamic window into your project's structural health. You get a practical way to spot complexity early, making your refactoring efforts precise rather than guesswork.

📰 https://news.ycombinator.com/item?id=49759501

🆔 @hackernewsgithubprojects

📦 programasweights/programasweights-python

ProgramAsWeights: Turn Text into Local Code

Turn plain English into tiny, runnable software that lives entirely on your computer. ProgramAsWeights takes a written description of a task and compiles it into a small neural function that runs offline without needing an internet connection or API keys. This is perfect for jobs like fixing broken data or sorting emails, where a full artificial intelligence is too heavy and simple text rules are too rigid. The coolest part is that you can describe what you want in normal words, and the tool handles the rest locally. No more sending your data to the cloud for simple tasks.

🆔 @hackernewsgithubprojects

📦 karanb192/claude-code-hooks

Claude Code Hooks: The Safety Net for AI Coding

Give your AI coding assistant a literal safety net that stops it before it breaks things. Claude Code Hooks is a collection of tiny scripts that sit between the AI and your computer, checking every single action it tries to take. The coolest part is that it runs outside the AI itself, so the robot cannot simply talk its way past the rules. It stops dangerous commands, protects your secret keys, and even blocks the AI from changing its own security settings. It is basically an invisible bouncer for your code, making sure your helpful assistant never accidentally deletes your work or leaks your passwords.

🆔 @hackernewsgithubprojects

📦 anmolkapil/plexo

Plexo: Download Faster with Multiple Networks

Plexo is the desktop app that finally uses all your internet connections at once to make downloads faster. Usually, your computer picks one connection and leaves the others sitting idle, but Plexo fixes that by breaking your file into small pieces and sending them out over Wi-Fi, Ethernet, and even a phone connection simultaneously. It stitches these pieces back together so you get the full file much quicker than any single line could manage. It is a clever way to squeeze every drop of bandwidth you already have, turning your setup into a speed boost without extra hardware.

🆔 @hackernewsgithubprojects

📦 ashhart/mcdma

MCDMA: Connect Your Mac to NVIDIA GPUs

MCDMA lets you wire your Mac directly into an NVIDIA GPU cluster, creating a raw data highway between Apple silicon and Linux hardware. It solves a messy problem where moving data between different computer chips usually requires slow, complicated software steps. By using a special network card, this tool moves large chunks of memory instantly, skipping the usual traffic jams. The real magic happens when you split a big AI task between machines, letting one system handle the heavy lifting while another manages the final output. It is experimental and needs specific hardware, but it shows how far we can push cross-platform speed.

🆔 @hackernewsgithubprojects

📦 cobanov/awesome-jev

Awesome Jev: Typed Decisions for AI

Build software that makes fast, confident decisions without generating endless text. Awesome Jev curates the entire ecosystem around Jev, a model from TypeSafe AI designed for typed, probabilistic choices like classification and routing. Instead of chatting, it takes your program's state and returns constrained answers with probabilities, letting your code handle the logic. The most surprising part? People are using it to pilot drones, rank coding agents, and even play Mario, proving it works way beyond simple data filters. It is not about replacing large language models, but about giving your apps a reliable, structured brain for high-speed logic.

🆔 @hackernewsgithubprojects

📦 expo/skills

Expo Skills: The Missing Brain for Your AI Coder

The Expo team built a cheat code for your AI coding assistant that teaches it how to actually build real mobile apps. This is not just another list of generic tips; it is a curated set of official instructions that tells your AI exactly how to handle the tricky parts of mobile development, like native animations, offline data fetching, and the specific quirks of iOS and Android. If you have ever watched your AI hallucinate a non-existent function or write web code that simply does not work on a phone, this is the fix.

🆔 @hackernewsgithubprojects

📦 0xsero/harness-bridge

Harness Bridge: One Tool for All Coding Agents

Harness Bridge is the local launcher that instantly connects any coding assistant to your own private model server without touching your existing files. The project solves the friction of manually editing configuration files or pasting API keys into command lines by handling the entire connection through environment variables that vanish when you are done. You simply pick a model from your local server, choose your favorite coding agent like Claude Code or Codex, and the tool opens a terminal already ready to work. This means your keys stay secure, your setup stays clean, and switching between different AI providers takes just one click.

🆔 @hackernewsgithubprojects

📦 dimentary/llm-robotics-playground

LLM Robotics Playground: AI Built These Robots

The LLM Robotics Playground is the collection that proves a frontier language model can engineer working robot controllers for delicate physical tasks. This library documents how a human guided an AI to build simulations where robots untangle headphone cables, carry objects on six legs, and even write code on a whiteboard with a marker. The most striking experiment features a robotic hand holding a pencil to draw a detailed dove, matching the intended path to within a fraction of a millimeter. It is a fascinating look at how language models are moving beyond text generation to actually solve complex, real-world mechanical problems.

🆔 @hackernewsgithubprojects

📦 jianglongye/dexterous-astra

Dexterous Astra: A Hand That Spins a Pen

A robot hand can spin a pen and twist a Rubik's cube. This project, called dexterous Astra, shows how simulated robot hands master these dexterous tricks. It provides ready-to-run environments and trained policies. You can run a zero-shot script that simply tells the hand to spin the pen. Or you can load a trained policy to see complex moves. The goal is to study how hands interact with objects. It is a clean way to test manipulation skills. You don't need to build the physics engine yourself. The code is set up for easy experimentation. This is a great resource for anyone studying robot hands. It makes complex dexterity accessible.

🆔 @hackernewsgithubprojects

📦 blader/humanizer

Humanizer: Make AI Text Sound Human

Take your AI generated prose and make it sound like a real person actually wrote it. Humanizer strips away the robotic habits that make text feel artificial, like forced lists or dramatic pauses, without changing your facts. It spots twenty five specific tells where language models slip up. The coolest part? It matches your unique voice if you paste a sample of your own writing. No more sounding like a chatbot. Just clean, natural prose that is actually yours. Give it a try.

📰 https://news.ycombinator.com/item?id=47362115

🆔 @hackernewsgithubprojects