📦 leggedrobotics/terrain-grounded-imagination

Terrain-Grounded Imagination: How Robots Learn to Walk

Teach a robot to walk over broken ground by letting it dream about the terrain. This project trains a humanoid to navigate stairs, gaps, and stepping stones entirely through a learned world model. Instead of hard-coding the ground into the robot’s brain, it reads the terrain directly from a heightmap, allowing the robot to imagine its next steps and adjust its balance based on exactly what it feels under its feet. This approach makes the robot’s perception of gaps and edges precise, ensuring it knows when a foot is unsupported.

🆔 @hackernewsgithubprojects

📦 jrryzh/f2g-pose

F2G-Pose: The One-Shot Object Scanner

F2G-Pose lets you figure out exactly where an object is in three-dimensional space from just a single photo and depth scan. You know how it feels when you snap a picture of a mug on your desk and your phone’s camera can’t quite tell which way it’s facing? This project fixes that. It takes a standard image and a depth map, and in one go, it guesses the object’s exact position, its size, and even what the whole shape looks like, even if you only saw half of it. No labels, no special templates, just a quick guess.

🆔 @hackernewsgithubprojects

📦 harunrrayhan/content-machine

Content Machine: Own Your Content Pipeline

Capture ideas straight from your phone and turn them into scheduled social posts without ever opening a separate app. That is exactly what Content Machine does. It acts as a private hub where you can jot down thoughts via a simple text note or even a voice memo. The system then uses AI to organize those messy fragments into clear, actionable content. Once your idea is ready, you can schedule it to publish automatically across your social channels. You get a single place to manage your creative workflow, keeping everything in sync. No more juggling multiple apps.

🆔 @hackernewsgithubprojects

📦 apolinario/urnas-tardias

Urnas Tardias: The 5,000 Votes That Never Should Have Counted

A new data analysis reveals exactly how many votes were cast after the legal deadline in Brazil's 2026 second round. The project, called Urnas Tardias, digs into the raw machine logs published by the electoral court to track every single vote processed after five in the afternoon. It turns out that only about five thousand votes were actually counted late across the entire country. This tiny fraction represents less than one percent of the massive lead secured by the winner. The analysis proves that extending voting hours had virtually no impact on the final outcome. The takeaway is simple: the election result is mathematically immune to late vote processing.

🆔 @hackernewsgithubprojects

📦 blagoysimandov/asd-ste100-writer-skill

The AI Skill That Writes Like a Pilot Manual

An AI coding assistant can now rewrite dense technical manuals into simple, plain English that non-native speakers and machines can easily understand. The ASD-STE-100 Writer Skill plugs directly into tools like Claude Code and GitHub Copilot to apply the strict writing rules used in aviation maintenance. It takes a complex paragraph full of long sentences and passive verbs and transforms it into short, direct commands. The result is text that removes ambiguity and makes safety warnings impossible to miss. It is a fascinating look at how we make technical data clearer.

📰 https://news.ycombinator.com/item?id=49108318

🆔 @hackernewsgithubprojects

📦 lebbitheplow/open-dungeon-master

Run D&D Locally With Open Dungeon Master

Run D&D 5e entirely on your own computer. Open Dungeon Master is a fully local app that generates the story and scene images right on your machine, so your campaign data never leaves your hands. The standout feature is the AI Dungeon Master who narrates the adventure, tracks initiative, and enforces the complex D&D 5e rules automatically, making it easy to run a session without needing a separate DM or online subscription. It handles character sheets, battle maps, and even voice output, giving you a complete, private tabletop experience.

🆔 @hackernewsgithubprojects

📦 datawhalechina/zero-to-sglang

Zero to SGLang: Build Your Own AI Engine

This free course lets you build a working AI server from absolute scratch before you ever touch the real code. It’s a complete roadmap where you hand-code a mini engine, adding features like memory optimization and request handling step by step. This is rare because most tutorials just show you a library without explaining how the inside works. By building it yourself, you actually learn what is happening under the hood, not just how to call a function. It is the perfect way to stop being a black-box user and start understanding the real logic behind modern AI speed.

🆔 @hackernewsgithubprojects

📦 cristianoc/tscert

TScert Proves Your Code Safe

TScert stops you from shipping TypeScript code that crashes at runtime by demanding mathematical proof before it accepts your function. Standard type checkers often miss bugs like reading undefined data, but this independent tool refuses to verify a function unless it can formally prove every step is safe using Lean. It doesn't just check types; it validates a logical certificate to guarantee your code behaves exactly as expected for every valid input. This means if TScert accepts your function, you can trust it won't break under the conditions you defined.

📰 https://news.ycombinator.com/item?id=50366465

🆔 @hackernewsgithubprojects

📦 jxy333967/jianxunyun

Jianxunyun: Your Guide to Stable Access

Keep your online access stable and private with Jianxunyun, a dedicated virtual private network service. Instead of guessing which server works, this repository serves as your personal cheat sheet for every official entry point. It lists the main websites and specific backup domains you need if the primary links get blocked. This is incredibly useful because digital services change addresses frequently, and having a static reference means you never lose your account or subscription data. You can switch between these addresses instantly to bypass regional restrictions. It is a simple, practical tool for anyone needing reliable, secure connectivity without the hassle of digging through forums.

🆔 @hackernewsgithubprojects

📦 sandflow-ai/sand3d

sand3d: 3D Modeling in Python for AI

sand3d is the 3D modeling library that lets AI build complex objects with surprisingly short code. It turns Blender into a text-based toolkit where a language model writes a few simple lines to create a whole machine, like a washing machine, in under forty lines. This is powerful because the entire instruction set fits inside an AI prompt, meaning a model can generate and explain a 3D design in plain, readable Python. It is not just for static shapes; it handles moving joints and parts, making it perfect for building robotic assets and game-ready models.

🆔 @hackernewsgithubprojects

📦 livliv-dev/player

LivLiv Player: Your Webcam Avatar

Drop an animated character onto your website with just a few lines of code. This is LivLiv Player, a tiny tool that brings two dimensional avatars to life right in your browser. The coolest part is that you can plug in your webcam, and the character mirrors your real-time facial expressions. It tracks your face locally on your device, so your avatar smiles, blinks, and talks just like you do. It is incredibly lightweight, with no heavy dependencies, making it perfect for adding a personal touch to stream overlays or interactive pages. No need for expensive software.

🆔 @hackernewsgithubprojects

📦 kingdaswinx/templum

Templum: Your AI Agents on One Canvas

Templum turns your computer screen into a giant digital whiteboard where you can pin down as many terminal windows as you want. You can drag them around, zoom out to see the big picture, or snap them together like a puzzle. The coolest part is that it acts as a bridge for your artificial intelligence coding tools. You can literally draw a cable from an AI agent to a to-do list or a note, allowing the robot to read your plan and start typing code into a terminal next to it. It is built with Rust and runs directly on your graphics card, so it feels snappy and native.

🆔 @hackernewsgithubprojects

📦 sawyerhood/do-computer

Do Computer: Your Own Personal AI Agent

Give your AI assistant a real computer to work on. Do Computer is an open-source project that lets you host your own always-on personal agent, called a Dood. Unlike standard chatbots that forget everything when you close the tab, this agent keeps running, remembers your preferences, and can even execute code on a machine you control. It is fascinating because it puts the power of a persistent, self-hosted AI worker directly in your hands without relying on a single corporate cloud. You get a customizable agent that learns skills, manages schedules, and connects to your favorite tools. It is your own digital partner, ready to work whenever you need it.

🆔 @hackernewsgithubprojects

📦 athei/mtld3d

Play Windows Games on Mac with Metal

mtld3d lets you run older Windows games directly on macOS by translating their legacy graphics code into Metal, Apple’s native language for high-speed rendering. It swaps out the standard compatibility layer with a custom engine that handles the heavy lifting right on your Mac’s hardware, bypassing the usual slowdowns. The most exciting part is how it runs these programs without needing a full virtual machine, keeping everything fast and responsive. You get true performance improvements rather than just running old software in a box. This project is a great example of how much further we can push cross-platform gaming.

🆔 @hackernewsgithubprojects

📦 kanapitsas/portal2-recomp

Portal 2 on iPhone

Play Portal 2 natively on your iPhone and Apple Silicon Mac without an emulator. This project rewrites the game’s original computer instructions into code your device runs directly, eliminating performance penalties from translation layers. The most surprising feature is firing portals by tapping the back of your phone, turning your device's physical shape into part of the puzzle-solving experience. It handles graphics and networking directly on your hardware, allowing smooth gameplay and local multiplayer with another phone or computer. You simply need your own copy of the game to build this. If you love classic puzzles, this is a fascinating look at running old games on modern hardware.

🆔 @hackernewsgithubprojects

📦 majorminorlabs/hermes-voice

Hermes Voice: Talk to Your AI Agent

Hermes Voice is the home integration that lets you talk to your personal AI agent using just your voice. Instead of typing into a chat box, you simply say Hey Hermes to an old Amazon Echo, and your custom agent responds out loud. It connects Home Assistant to your specific AI profile, meaning the assistant remembers your preferences and keeps the conversation flowing naturally. You can even set it up so your agent controls your lights or searches the web while you speak.

📰 https://news.ycombinator.com/item?id=50381927

🆔 @hackernewsgithubprojects

📦 cvg/flexglue

FlexGlue: The Image Matcher That Knows When to Stop

FlexGlue is the image matching model that knows when two pictures are actually the same place. It solves a tricky problem where look-alike buildings or repeated patterns fool standard software into thinking they match when they do not. Instead of blindly connecting points, this tool predicts if the images show the same scene and stops working early if they do not. This saves time and prevents false connections. It handles tricky lighting and angles with ease. You get a cleaner result without the extra noise. This is a smart way to handle complex visual data. It is built for real world reliability. Trust your results more.

🆔 @hackernewsgithubprojects

📦 tradertanmay/ai-agents-zero-to-hero

AI Agents: Zero to Hero

AI Agents: Zero to Hero is the Python curriculum that strips away framework magic to show you exactly how autonomous agents work from first principles. It replaces opaque black-box tools with a clear, code-first path that walks you through the observe-decide-act loop using only standard Python. The surprising strength here is its zero-dependency approach, meaning you can build, run, and debug a complete agent loop without installing heavy packages or managing API keys. This gives you a deep, framework-independent understanding of memory, tool calling, and safety before touching real production systems. It is the clearest way to move beyond buzzwords and actually build intelligent software you can trust.

📰 https://news.ycombinator.com/item?id=49973196

🆔 @hackernewsgithubprojects

📦 tiptreesystems/lacuna-research-mcp

Lacuna Research MCP

Lacuna Research MCP is the knowledge bridge that turns your coding agent into a full-time research assistant. It plugs into your favorite editor to search through millions of machine learning papers, authors, and venues. The coolest part is that it traces code repositories back to their original academic papers, so you know exactly where every tool came from. It also surfaces AI generated research proposals, giving you fresh ideas before you start coding. It is a map of the entire field, making it easy to find prior work and avoid reinventing the wheel.

🆔 @hackernewsgithubprojects

📦 giveen/decomp.anything

Turn Any Binary Into Readable Code

Reverse engineer any software by letting an AI agent write C code that compiles to the exact same bytes as the original program. The tool splits a binary into small pieces, then has the agent guess the source code for each part. It uses the original compiler to build the guess and compares the output byte by byte until they match perfectly. If they do not match, the agent sees the difference and tries again. This works for old games, Windows apps, and Linux tools. It runs entirely on your own computer, so your files never leave your machine. No account needed, no cloud, no privacy trade-offs.

🆔 @hackernewsgithubprojects