📦 streameval/streameval

StreamEval: The First Test for Live Video AI

StreamEval is the first benchmark ever built to test AI systems that generate live, interactive video in real time. Instead of judging a finished clip, it feeds instructions to a model one by one while it is already running, measuring how well the system adapts to new requests without losing track of the scene. This matters because most current video tools create static files, but the future of virtual avatars and live worlds depends on software that can think and update on the fly. By standardizing this difficult test, researchers finally have a fair way to compare who is actually building usable live video technology.

🆔 @hackernewsgithubprojects

📦 showlab/awesome-multimodal-embodied-agent

How AI Robots Learn to Think

This collection tracks how artificial intelligence moves from clicking buttons on your screen to physically moving objects in the real world. It organizes research on embodied agents, which are systems that combine seeing, planning, and acting to solve tasks in physical spaces. The project uses a framework called PAPAV to categorize how these agents perceive their environment, anticipate outcomes, plan moves, execute actions, and verify results. This matters because it shows the step-by-step progress of making machines that understand and interact with the world around them. If you want to know how software is becoming hardware, this is the map.

🆔 @hackernewsgithubprojects

📦 thesecuredanalyst/security-suite

Security Suite: AI-Powered Security Analysis

Security Suite is the open-source security platform that finally connects your scattered scan results into a single, actionable story. It handles the heavy lifting of digging up domain details, hunting for web vulnerabilities, and testing your application interfaces. But the real magic happens when you let it think. Using local or cloud-based artificial intelligence, it reads your raw findings and writes you a clear, executive summary. No more staring at a wall of technical noise. You get a plain-English explanation of what actually matters and how to fix it. It is like hiring a senior analyst who does not charge a cent.

🆔 @hackernewsgithubprojects

📦 unjs/jup

jup: The end of version mismatch headaches

Run the exact tool your project needs without manually installing it. This small program handles the right package manager or runtime for every job, checking that the download is safe before it runs. It saves the version on your computer so you can work offline, ensuring everyone uses the same setup. It keeps your normal commands working while removing the guesswork, making it a practical tool for developers who want consistent, verified environments without the usual complexity.

🆔 @hackernewsgithubprojects

📦 planetis-m/skills

Skills: Teaching AI Nim Coding Rules

Give your AI assistant verified coding skills for the Nim programming language. The repository called Skills solves a specific problem: large language models often hallucinate or use outdated syntax, but this collection provides curated, empirically tested modules that guide the AI on exactly how to handle Nim specifics. Think of it as a cheat sheet for your bot, covering everything from memory management and API design to debugging and testing. It works by simply linking these skill folders into your agent directory, so the AI automatically applies the correct rules for tasks like C bindings or error handling.

🆔 @hackernewsgithubprojects

📦 thebcd69/bcd-games

Run PC Games on iPhone

BCD lets you play Windows PC games directly on your iPhone and iPad, with Steam and GOG support built right into the app. It uses Wine and FEX to run the code on Apple’s chips, translating it so it works smoothly without a computer. The coolest part is that it handles graphics by converting Windows Direct3D calls into Metal, the system graphics language for Apple devices. This means you can launch big titles like God of War or GTA V straight from your library, complete with controller support and touch controls.

🆔 @hackernewsgithubprojects

📦 code-zm/plex2shell

Plex2Shell: The Critical Plex Vulnerability

Plex2Shell is the security research tool that exposes a critical, unauthenticated remote code execution flaw in older versions of Plex Media Server. It chains four distinct security bugs together, allowing anyone on the internet to take full control of a server without needing a user account or password. This is dangerous because it bypasses all standard access controls, turning a simple media player into a potential open door for attackers. The project serves as a stark warning for server administrators to check their patch levels immediately. If you run Plex, this is your cue to update right now, because leaving that door open is how you get hacked.

🆔 @hackernewsgithubprojects

📦 agnesai-labs/agnes-harness

Agnes Harness: The AI Agent That Remembers

Agnes Harness is the agent harness that turns every messy deployment into reusable muscle memory. When engineers build AI tools for specific customers, they usually start from scratch every time, losing all that hard-won context. This project fixes that by keeping a persistent memory of how tasks actually work. It acts like a reliable coworker who remembers exactly which buttons to press and which approvals to ask for. Instead of forgetting everything after the first job, it saves the method so your next project starts smarter. It is a developer preview, meaning it is raw but ready for builders.

🆔 @hackernewsgithubprojects