📦 end2end-diffusion/diffusion-bench

The All-in-One Benchmark for Generative Diffusion Models

Standardize your generative diffusion research with a unified codebase designed for holistic model evaluation. This project enables researchers to train and evaluate various diffusion transformers across both ImageNet and text-to-image tasks using a single, consistent interface. You can leverage extensive support for latent space methods like VAEs and RAEs, alongside diverse transport mechanisms and loss functions. The repository provides integrated tools for offline evaluation, including metrics like FID, IS, and VQAScore, helping you move beyond simple ImageNet benchmarks. Explore this repository to streamline your experimentation and contribute to a more robust evaluation standard for modern diffusion models.

📰 https://news.ycombinator.com/item?id=48654274

🆔 @hackernewsgithubprojects

📦 miaai-lab/deepseek-v4-flash-dual-dgx-spark-1m-context

Deploy DeepSeek V4 Flash on Dual DGX Spark Nodes

Deploy the DeepSeek V4 Flash reasoning model across two NVIDIA DGX Spark nodes to unlock powerful distributed inference capabilities. This repository provides a streamlined setup using Docker and shell scripts to orchestrate a vLLM server across your cluster. It leverages InfiniBand for fast inter-node communication, utilizes FP8 KV-cache for memory efficiency, and supports a massive 1 million token context length for complex tasks. By implementing tensor parallelism and multi-token prediction, you can maximize your hardware performance for high-end reasoning and coding workloads. Use this framework to scale your AI infrastructure and achieve high-performance model serving today.

🆔 @hackernewsgithubprojects

📦 facebookresearch/lamp

Track 3D Human Motion with Multi-Camera Headsets

Track human motion in a 3D world using egocentric multi-camera headsets by leveraging a unique lift-then-fit approach. This system disentangles observer and target movement by lifting 2D body keypoints from multiple camera angles into a unified 3D reference frame. By fitting motion directly to this 3D ray cloud using a specialized spatio-temporal transformer, the project achieves state-of-the-art accuracy for egocentric tracking. It is a powerful tool for researchers and developers working with AR and VR human pose estimation, providing a robust solution for capturing natural, metric-scale human movement from wearable devices in real-world environments.

🆔 @hackernewsgithubprojects

📦 elhart05/airmongui

Simplify Wireless Auditing with AirmonGUI

Manage your wireless auditing tasks by using this web-based interface that centralizes the powerful aircrack-ng toolkit. Instead of juggling multiple command-line flags, this project provides a guided workflow for enabling monitor mode, scanning for networks, capturing WPA handshakes, and performing deauthentication attacks. The tool includes a live table for network analysis, a PTY terminal for raw command access, and comprehensive log tracking to keep your session state organized. By running locally, it gives you a clean visual dashboard for security testing on networks you control, making complex wireless diagnostics feel straightforward and efficient for your lab assessments.

🆔 @hackernewsgithubprojects

📦 seeed-projects/co-create_esp-fly

Build Your Own Micro Drone with ESP-FLY

ESP-FLY is the DIY micro drone kit that finally makes aerial robotics accessible through the power of the Seeed Studio XIAO ESP32-S3. This project provides a complete open-source platform for assembling and flying a compact fifty-millimeter quadcopter using a custom flight controller board with integrated six-axis motion sensing. You can pilot the drone either via a direct smartphone Wi-Fi connection or through a radio controller using ESP-NOW for precise flight. It is an excellent hands-on way to explore embedded systems and flight stabilization, offering a fun and educational path to mastering your own custom-built micro drone.

🆔 @hackernewsgithubprojects

📦 adshao/flounder

Automate Security Audits with Flounder

Automate your security audit workflow by using this tool to turn coding agents into end-to-end security auditors. It allows you to provide a project link, repository, or address, letting the system autonomously prepare the workspace, map the attack surface, and perform deep audits. By sandboxing all model-generated code and execution, it ensures that findings are grounded in evidence by requiring successful local proof tests before reporting them. Whether investigating incidents or performing bounty audits, it creates a structured, framework-agnostic pipeline that keeps your audit process secure and evidence-backed. Give it a try to streamline your security research.

📰 https://news.ycombinator.com/item?id=48657652

🆔 @hackernewsgithubprojects

📦 serveurpersocom/qwentts.cpp

Run Local AI Text-to-Speech with Qwen3-TTS

This project provides a robust C++17 implementation of Qwen3-TTS, enabling high-quality, local AI text-to-speech synthesis on standard hardware. By leveraging the GGML library, the software runs efficiently across CPU, CUDA, ROCm, Metal, and Vulkan backends. It offers advanced capabilities like zero-shot voice cloning, named speaker selection, and flexible voice design through simple text instructions. With support for eleven languages and streaming audio output, it allows developers to integrate powerful, low-latency speech generation into local applications. This streamlined toolchain makes sophisticated, multi-stage neural audio synthesis accessible and portable, ensuring consistent performance without relying on cloud infrastructure for your speech synthesis needs.

🆔 @hackernewsgithubprojects

📦 cortexprism/cortex

Build and Manage AI Agents with Cortex

Orchestrate complex AI agent workflows and automate tasks directly through an integrated open-source operating system. This platform provides a centralized environment to manage autonomous agents, complete with tools for code execution, web research, and persistent memory storage. It streamlines agent development by offering built-in observability, security guardrails, and diverse integrations with messaging channels like Slack and Discord. Whether you are building sophisticated sub-agent structures or automating development pipelines, this system offers the modular architecture needed to deploy functional AI agents effectively. Explore its capability to transform how you build, monitor, and interact with autonomous digital assistants.

🆔 @hackernewsgithubprojects

📦 velobase/velobase-harness

Launch Your AI SaaS Fast with This Open Source Boilerplate

Build and monetize AI applications rapidly using a comprehensive open-source SaaS boilerplate designed for the modern web. This project provides a complete foundation by integrating essential features like Stripe billing, usage-based pricing models, affiliate tracking, and robust authentication right out of the box. Built with the T3 stack and Next.js, it simplifies the transition from initial code to a revenue-generating platform. You get access to pre-configured modules for AI chat, payment processing, and user management, allowing you to focus purely on your product features. Save significant development time and get your software running in production today.

🆔 @hackernewsgithubprojects

📦 per-simmons/unreal-agent-harness

AI Agents Building Cities in Unreal Engine 5.8

Unreal Agent Harness is the experimental platform that empowers AI agents to autonomously build and refine complex 3D environments directly within Unreal Engine 5.8. By leveraging the official Unreal MCP, the project provides a sophisticated bridge for an agent to gain visual perception through automated viewport capture and tactile control over scene assembly. It enables advanced workflows like procedural city generation, photorealistic terrain streaming via Cesium, and automated QA loops where the agent acts, inspects, and corrects its own work. This harness turns the engine into a programmatic canvas, making it an essential tool for those exploring AI-driven game development.

🆔 @hackernewsgithubprojects

📦 toyotaresearchinstitute/turbompc

TurboMPC: High-Speed GPU Model Predictive Control

TurboMPC delivers high-speed, scalable, and differentiable model predictive control directly on the GPU, significantly accelerating complex robotic planning tasks. By utilizing sequential quadratic programming and the Alternating Direction Method of Multipliers, this repository enables efficient, batchable solves that are essential for real-time applications like drone navigation and space systems. It leverages high-performance linear system solvers, including NVIDIA cuDSS, to maximize throughput and allow for seamless integration with JAX for gradient-based optimization. Developers can easily switch between CPU and GPU backends to balance flexibility and raw power, making it a robust tool for advanced trajectory optimization and control research.

🆔 @hackernewsgithubprojects

📦 opendatabox/memorydata

Benchmark Memory-Augmented AI Agents with MemoryData

MemoryData provides a unified execution interface for benchmarking memory-augmented AI agents, enabling direct comparison across twenty-two distinct method presets. Previously, evaluating these architectures required disparate runtimes and custom harness implementations, making reliable performance metrics difficult to track. This suite solves that problem by integrating four major benchmark families into a single pipeline, allowing researchers to standardize testing across diverse paradigms including sequential, topological, and hybrid structures. Users can execute experiments via a central launcher, generating structured result artifacts and persistent agent states for consistent analysis. This repository ultimately provides a stable foundation for advancing memory-augmented agent research.

🆔 @hackernewsgithubprojects

📦 ggjy/awesome-agent-engineering

Understanding the Foundation of AI Agent Engineering

Large language models are shifting from simple question answering to active task completion, a transition that fundamentally relies on execution harnesses. This repository offers an organized collection of academic papers, systems, and benchmarks that define the agent as the sum of a foundation model and a runtime infrastructure. It identifies six key harness responsibilities, including observation interfaces, control loops, and verification systems, which determine how agents perceive environments and recover from errors. By exploring four engineering paradigms ranging from prompt design to agent-native training, these resources provide a comprehensive framework for building reliable, autonomous, goal-directed systems.

🆔 @hackernewsgithubprojects

📦 astroautomata/thousandworlds

ThousandWorlds: A Climate Emulation Benchmark

ThousandWorlds is the climate emulation benchmark that simplifies predicting exoplanet atmospheres for habitability studies. It offers seventeen hundred sixty simulations across five general circulation models, capturing complex variables like temperature and wind speeds on a structured grid. Researchers can leverage its nested datasets and baseline methods to train and evaluate models against standardized physics-based metrics. Whether you are performing parameter-to-field regression or assessing inter-model uncertainty, this repository provides the necessary tools and precomputed results to advance your work. Download the dataset and start testing your climate emulation models on this robust, expert-validated platform today.

🆔 @hackernewsgithubprojects

📦 orange-3dv-team/car

CaR: Revolutionizing Video World Model Memory

CaR is the video generation framework that finally solves long-term memory challenges in world models through efficient implicit retrieval. This project addresses the difficulty of maintaining visual consistency across complex camera movements by using an attention-driven mechanism to retrieve historical context. By combining a lightweight compression network with specialized retrieval attention, it enables smooth scene exploration and extended video synthesis from single images. The team also introduces the SceneFly dataset, which provides extensive synthetic video data for training these models. This work effectively bridges the gap between static generations and dynamic, consistent video environments for future simulations.

🆔 @hackernewsgithubprojects

📦 spacepxl/comfyui-vae-utils

Unlock Advanced VAE Control in ComfyUI

ComfyUI-VAE-Utils is the custom node toolkit that finally enables advanced VAE loading and processing options unsupported by standard ComfyUI. This repository provides dedicated nodes for loading VAEs with automatic input and output channel detection, specifically tailored for modern models like Wan. You gain access to specialized decoding capabilities, including automatic upscaling and tiling, alongside a neural latent upscaler that significantly outperforms basic interpolation methods. By bridging these functional gaps, the project simplifies complex generation workflows, allowing you to effectively manage latent manipulation and high-quality output decoding directly within your existing pipeline for better image and video results.

🆔 @hackernewsgithubprojects

📦 david-magdy/cvil

Master Computer Vision Interviews with This Checklist

Prepare for computer vision and machine learning interviews by using a structured, phase-based map designed to track your knowledge from core concepts to advanced production techniques. This repository provides a clear, organized list of essential topics, covering everything from fundamental mathematics, statistics, and CNN architecture to complex subjects like vision transformers, object detection, and tracking systems. It emphasizes developing a deep intuition for why specific models exist and the tradeoffs they introduce, helping you bridge the gap between academic theory and practical application. Leverage this checklist to systematically identify your study gaps and confidently master your next technical interview.

🆔 @hackernewsgithubprojects

📦 h-freax/termhub

Build Your Developer Portfolio in Minutes with TermHub

TermHub is the terminal-style portfolio template that turns your resume into a professional developer website without writing code. It simplifies the entire process by using a clever pipeline where you provide your resume to an AI, which then generates the necessary markdown and configuration files for your site. The project features a sleek, responsive design with a classic terminal aesthetic and supports bilingual content. You can even use its built-in server tools to automate site creation, allowing you to manage publications, projects, and work history simply by editing text files. It is the fastest way to showcase your work online.

📰 https://news.ycombinator.com/item?id=47378150

🆔 @hackernewsgithubprojects

📦 xingzhejun/d-opsd-code

Understanding d-OPSD for Diffusion Large Language Models

d-OPSD is the specialized training framework that introduces on-policy self-distillation for diffusion large language models. This project provides the essential implementation code and environmental setup required to train and evaluate models using this new approach. By modifying standard trainer files to enable specific batch generation configurations, it allows researchers to effectively apply self-distillation techniques across tasks like math and logic puzzles. The repository includes clear shell scripts for seamless training and performance evaluation, offering a structured way to improve model reasoning capabilities. This framework is a valuable resource for anyone working on optimizing advanced generative language models.

🆔 @hackernewsgithubprojects

📦 trickygo/moonsplat

MoonSplat: Real-Time 3D Reconstruction from Monocular Video

MoonSplat enables robust real-time 3D reconstruction and camera tracking using only standard monocular video sequences. By integrating global Sim(3) optimization with voxelized Gaussian splatting, the framework addresses common challenges like fragile pose estimation and optimization inefficiency in long-sequence scenarios. The implementation includes a color residual learning strategy to accelerate convergence and boost rendering quality, making it a powerful tool for applications like robotics and virtual reality. Users can configure parameters to balance speed and performance while processing custom data. This research project serves as an efficient, high-fidelity alternative for capturing dense 3D scenes from simple camera inputs.

🆔 @hackernewsgithubprojects