📦 zeyu-zheng/far

FAR Automates Mathematical Conjecture Review

Save hours of tedious literature digging with FAR, a new tool from CMU and Anysphere. It acts as an automated research assistant that scans massive paper archives to hunt down unsolved math problems matching your specific interests. Once it finds those open questions, it uses advanced AI agents to attempt proofs and strictly grades their validity. The system filters out the noise and only hands you the most promising, verified results for your expert eye. It turns a chaotic search process into a clean, reliable pipeline that highlights high-impact mathematical discoveries ready for human review.

🆔 @hackernewsgithubprojects

📦 francois141/upal

UPAL Finds Points and Lines at Once

UPAL is the computer vision tool that finally detects points and lines in a single pass. Most image matching systems handle corners or straight edges as separate problems, but this new research from ECCV 2026 combines them into one unified network. It identifies precise sub-pixel keypoints and their descriptors while simultaneously predicting a distance field that highlights straight lines. This allows the system to match features across different photos of the same scene, linking both individual points and long segments with high accuracy. By treating the whole scene as a connected map rather than isolated parts, it helps computers understand spatial relationships much better.

🆔 @hackernewsgithubprojects

📦 facebookresearch/moe_vie

MoE-ViE: The Efficient Vision Encoder

MoE-ViE achieves high accuracy on complex video and image tasks while using a specialized mixture of experts architecture to keep computations efficient. This new vision encoder from Meta Research handles understanding both photos and video clips by routing different parts of the visual data to specialized experts within a single transformer. It solves the problem of high computational cost in standard vision models, making it possible to process large-scale visual data without the heavy energy or hardware requirements usually needed. The repository provides the full model definition and a reproducible evaluation suite, so you can test its performance on standard benchmarks yourself.

🆔 @hackernewsgithubprojects

📦 wz-cs/vggt-align

VGGT-Align Fixes the Scale Drift Problem in 3D Maps

VGGT-Align is the 3D reconstruction framework that finally stops long videos from stretching and warping as they are mapped. When you try to build a 3D model from a long sequence of images, standard tools often suffer from scale drift. This means the map slowly gets bigger or smaller over time, causing the final 3D scene to look distorted and misaligned. The key insight here is that in places like driving scenes, certain geometric features stay the same size no matter which part of the video you look at. VGGT-Align uses these stable features to lock the scale in place.

🆔 @hackernewsgithubprojects

📦 cnulab/logico

LogiCo Finds Hidden Factory Defects

LogiCo is the anomaly detection framework that finally spots the weird stuff in factory photos. Most quality checks look for specific, known errors, like a scratch on a bottle. But LogiCo takes a different approach. It learns what a perfect object looks like, then flags anything that doesn't quite fit the pattern, whether it is a logical mismatch or a structural flaw. It uses a powerful vision model to understand complex images deeply, catching subtle defects that simpler tools miss. This is huge for manufacturing because it means you can detect problems you didn't even know were possible.

🆔 @hackernewsgithubprojects

📦 plaguekind/comfyui-plaguekind-nodes

PlagueKind Nodes: Fix Your Image Masks in ComfyUI

Stop fighting with misaligned masks in ComfyUI. PlagueKind Nodes solves the headache of resizing images and their masks separately, which usually leads to weird stretching or off-center crops. This single node ensures both the picture and the mask follow the exact same geometric rules, keeping your inpainting perfectly aligned. It handles tricky aspects like maintaining aspect ratios and ensuring dimensions fit specific model requirements without breaking your workflow. Plus, it lets you drag a crop box directly on the screen to visually select exactly where you want to cut, saving you from doing any manual pixel math. If your edits keep drifting out of sync, this is the fix.

🆔 @hackernewsgithubprojects

📦 dream-num/dsh-univer-office

Turn Natural Language Into Office Files With DSH Univer Office

Imagine asking an AI to build a complex spreadsheet or presentation deck and having it actually do the work, right in front of you. DSH Univer Office connects that power to DeepSeek Harness, letting you create, edit, and inspect spreadsheets, documents, and slides just by describing what you need in plain English. The coolest part is the live preview: as the AI builds your file, you can watch the changes happen in a window right next to your chat. It handles everything from calculating formulas to laying out slide pages, and even checks for messy text overflow so your work looks clean.

🆔 @hackernewsgithubprojects

📦 azrtydxb/procoder

Procoder: The Senior Developer Guardrail for AI

Procoder acts as a strict senior developer for your AI coding agents, refusing to let them commit code that is actually broken. It works by installing a hard commit gate that treats any unchecked tool or failed lint as a failure, ensuring no unfinished work slips through. The binary itself never touches your files, but it blocks the commit until the agent fixes every issue, from merge conflict markers to junk files. It even remembers past mistakes to prevent the same bugs from recurring, forcing the agent to act like a professional engineer instead of a guessing game.

🆔 @hackernewsgithubprojects

📦 andysalerno/r9700-serving

Run AI Locally on R9700s

Serve a large language model locally using two AMD R9700 graphics cards. This project provides the exact Docker files and configuration needed to run VLLM on this specific hardware, bypassing the usual setup headaches. It includes benchmarks proving the system can handle a twenty-seven billion parameter model with impressive speed. The repository tracks different software versions to show what actually works on your machine. If you have R9700s and want to run AI privately, this is the blueprint you need to get started.

🆔 @hackernewsgithubprojects

📦 furkankly/rataflow

Rataflow: Node Editors in the Terminal

Rataflow brings a fully interactive node-based editor to your terminal, letting you drag boxes and wire them together using your mouse without leaving the command line. Normally, terminal apps are limited to static text grids, but this library solves that by handling complex mouse interactions like panning, zooming, and connecting nodes with smooth animations. It works by rendering to a grid of characters, making it perfect for building visual tools that feel like modern desktop apps but run anywhere you can open a console.

📰 https://news.ycombinator.com/item?id=49345146

🆔 @hackernewsgithubprojects

📦 yukonga/ghostlock-app

GhostLock: One Tap to Root

A single tap on your phone can grant full system privileges without a traditional root manager. GhostLock is an Android app that exploits a race condition in the Linux kernel to escalate from a normal app to root access, bypassing standard permission boundaries. It works by running a native binary that races two CPU cores, effectively forcing the system to hand over administrative control. This is strictly for supported devices with specific kernel versions, and you need KernelSU or ReSukiSU installed to load the necessary modules.

🆔 @hackernewsgithubprojects

📦 zachahn/vomit

Vomit Fixes Claude's Garbled Speech

Vomit turns Claude’s garbled token spam into plain English using a separate local model. It works entirely on your machine, so your data never leaves your computer. When Claude tries to speak, this tool catches the output, pipes it to a local language model, and rewrites it into clear, human-readable text. You can run it as a background hook or use it to translate old session logs. It’s a quirky, practical solution for anyone finding the default output too noisy. If you run local LLMs, this is the cleanest way to keep your focus on the actual code rather than decoding robot-speak.

📰 https://news.ycombinator.com/item?id=49375996

🆔 @hackernewsgithubprojects

📦 ant-research/4danyone

Turn One Phone Clip Into 4D with 4DAnyone

4DAnyone is the video generator that takes a single casual phone clip and turns it into a fully explorable four-dimensional character. You do not need a special camera rig or expensive studio setup to capture someone moving in three dimensions. This tool accepts one standard video of a person and predicts how they look from every possible angle, including views from above, behind, and around them. It handles the tricky work of matching the motion and lighting so the generated angles look consistent. Once you have those multiple views, you can feed them into standard reconstruction tools to build a realistic digital twin.

🆔 @hackernewsgithubprojects

📦 fireredteam/fireredtts3

FireRedTTS3: Design Any Voice With Words

FireRedTTS3 lets you create a completely new human voice using only a text description, with no reference audio required. This open source project unifies voice cloning, voice design, and speech editing into one system. The surprising capability is its instruction-guided voice design, where you can describe traits like gender, age, timbre, and emotion in plain language to generate a unique speaker. It supports twenty four languages and twenty one Chinese dialects, making it highly versatile for global applications. By treating speech as a semantic representation, it allows for precise control over both the sound and the meaning of the audio.

🆔 @hackernewsgithubprojects

📦 google-research/envharness

Make AI Training Environments Dynamic

Turn frozen test environments into dynamic training grounds that adapt to an agent's specific weaknesses. Traditional benchmarks stay static, so they stop helping once the agent solves them. EnvHarness wraps existing benchmarks with plug-in components that reshape the starting state and interaction rules without touching the internal code. This allows a designer agent to diagnose flaws and generate targeted challenges in real time. The result is that skills learned in these dynamic worlds transfer back to the original benchmarks, boosting success rates and reducing the steps needed to solve tasks.

🆔 @hackernewsgithubprojects

📦 albertobarnabo/lazy-cat

How Lazy-Cut Saves 178x in AI Coding Costs

Imagine your AI assistant is a brilliant engineer who never sleeps and never asks questions. It just starts building. The problem is that this thoroughness often leads to massive waste, where simple tasks consume thousands of unnecessary tokens. A new project called lazy-cat changes this by teaching AI agents to be productively lazy. It forces the AI to pause and ask if there is a smarter, cheaper way to solve the problem before it starts writing code. In one test, generating five hundred user profiles took sixty-six thousand tokens without this tool, but only three hundred and seventy-two tokens with it.

📰 https://news.ycombinator.com/item?id=48450852

🆔 @hackernewsgithubprojects

📦 arvids-unavailable/opengym

Own Your Gym Data with openGym

Your fitness data currently lives in a corporate database, but openGym changes that by letting you run the entire app on your own hardware. This self-hosted tracker stores your workout history and body weight as simple files on your machine, meaning nobody else has access to your progress. It features a massive exercise library with animated demos and uses passkeys for secure, password-free login. The best part is that it works offline and syncs across your devices without any subscription fees or ads. You back up everything with a single folder copy, ensuring your training history is truly yours forever.

🆔 @hackernewsgithubprojects

📦 cyanchemistvoid/jiocinema-apex

JioCinema Apex: The Offline Streaming Toolkit

JioCinema Apex lets you download your favorite streams and watch them without a single second of buffering, even in a dead zone. It acts as a toolkit that handles offline mode, queue management, and recommendation engine access all in one package. You can save content for offline viewing on any device you own, and it keeps your progress synced across multiple screens at once. The setup is straightforward: grab the package, unzip it, and run the main file to start streaming. It is completely free to use, making it a handy tool for anyone who wants to control their viewing experience without relying solely on a constant internet connection.

🆔 @hackernewsgithubprojects

📦 mrbizarro/phosphene

Phosphene: Generate Videos and Voices Entirely on Your Mac

Phosphene turns your Mac into a complete local video studio, generating synchronized audio and video clips without a single cloud API call. You can train a custom digital character using just a handful of photos, then render that specific face speaking your lines in about six minutes on high-end Apple Silicon. This solves the major privacy headache of uploading sensitive creative assets to remote servers, keeping your entire workflow strictly on-device. It is the first tool to make character-consistent video generation as simple as a one-click install, proving that powerful AI media creation does not require expensive GPU clusters.

🆔 @hackernewsgithubprojects

📦 robotis-git/ai_sapiens

AI Sapiens: The Open Humanoid Bridge

AI Sapiens is the open humanoid platform that brings trained AI models directly to physical hardware. This repository from Robotis provides the essential ROS 2 tools to run neural policies on their K1 robot. It solves the difficult problem of transferring simulation logic into the real world. The sim-to-real bridge handles everything from safe startup checks to real-time joint control. It ensures your robot moves precisely and safely. You can switch between manual teleop and automated AI behavior with a single command. This setup makes it easy to test new movements without breaking your hardware. It is the practical foundation for anyone building physical AI.

🆔 @hackernewsgithubprojects