📦 redai-studio/relax

Relax: The Async RL Engine

Relax runs massive AI model training and inference on separate hardware clusters that talk to each other in real time, meaning the system never sits idle waiting for data. This open source engine from Xiaohongshu was built specifically to teach multimodal models, allowing them to learn from text, images, video, and audio simultaneously without needing a new framework for every single type of data. By splitting the heavy lifting of learning from the heavy lifting of answering, the whole pipeline stays busy and efficient even at scale. If you are working on complex AI training, this is the kind of architecture that keeps your expensive computer chips actually doing work.

🆔 @hackernewsgithubprojects

📦 harvard-cns/orla

Orla: The Smart Proxy for AI Agents

Orla is the execution engine that stops you from hard-coding which model handles each part of your AI workflow. Built by researchers at Harvard, this Go-based daemon sits between your agent and the backend, acting like a smart traffic cop. You simply tag requests by their purpose, and Orla routes them to the best available option. The magic happens in the background. An external optimizer watches your real-world performance data and constantly retunes the routing policy. It might swap to a cheaper model for easy tasks or escalate to a powerful one for complex jobs. Your code never changes. It just runs faster and costs less.

🆔 @hackernewsgithubprojects

📦 anton-barinov/php2shell-patches

PHP2Shell Patches: Fix Old PHP Safely

PHP2Shell patches is the emergency toolkit that actually keeps your legacy servers safe without breaking your file uploads. While the official fix for this critical security hole simply disables progress bars to stop the exploit, this repository takes a smarter approach by fixing the root cause directly. It lets you patch old PHP versions like 5.6 and 7.4 so your forms and upload trackers keep working exactly as they always have. This is the definitive fix for anyone running outdated software who refuses to pick between security and functionality. Apply these specific patches now to secure your server while keeping everything running smoothly.

🆔 @hackernewsgithubprojects

📦 recogardtech/autopilotpm

AutoPilotPM: AI Trading Across 1000+ Markets

A single open-source project can watch over a thousand different betting and crypto exchanges at once, finding gaps in price and placing trades before you even blink. AutoPilotPM is an autonomous system that uses artificial intelligence to scan markets like Polymarket, Kalshi, and Binance, hunting for opportunities where prices disagree across different platforms. It doesn't just find these gaps; it automatically executes the trades and manages your risk limits so you don't have to sit glued to a screen. Think of it as a tireless, digital trader that works around the clock to lock in small profits from market inefficiencies.

🆔 @hackernewsgithubprojects

📦 indras-mirror/xolotl

Xolotl: A 552B Model on One PC

Xolotl is the local inference engine that runs a 552 billion parameter frontier model on a single gaming PC. Most people assume that massive artificial intelligence requires a server farm, but this project proves you can run the DeepSeek V4.1 Flash model on just one graphics card and standard computer memory. It achieves this by cleverly storing the model's weights across different hardware layers and pulling up only the pieces needed for each word. This means you get real coding capabilities and tool calling locally, without paying API fees or waiting for cloud latency. If you want powerful AI on your own hardware, this is the one to check out.

🆔 @hackernewsgithubprojects

📦 ciscobugger/tacit

Tacit: The OS That Actually Looks Like One Thing

Tacit is the operating system that finally makes your laptop feel like a single, cohesive object instead of a patchwork of random parts. Running on Arch Linux, it ditches the standard taskbar for a sleek notch at the top of your screen that quietly handles notifications and status updates without getting in the way. Every single piece, from the file manager called Pinax to the ad-blocking browser named Thyris, is designed to fit together with a unique visual language. It is not just a skin; it is a complete environment where the fonts, icons, and apps all speak the same visual dialect.

🆔 @hackernewsgithubprojects

📦 ltmoerdani/opencode-copilot-chat

Hack VS Code to Use Free AI Models

Stop paying for premium coding subscriptions when you can just grab the good stuff for free. The OpenCode Copilot Chat extension lets you plug your own API keys directly into VS Code, unlocking over thirty frontier AI models without a single monthly bill. Think DeepSeek, Kimi, GLM, and even GPT and Claude, all sitting right in your native Copilot chat picker. The coolest part is that it keeps the native tool calling and agent mode working perfectly, so your AI can still read files and run commands just like the paid version. You get the full power of these heavy hitters without the subscription lock in.

🆔 @hackernewsgithubprojects

📦 rosehahah/the-machine

The Machine is Now on Your Mac

This open source project puts the AI surveillance system from Person of Interest directly on your Mac. It uses your camera to box every person it sees, recognizes you as the Admin, and detects if someone picks up a weapon or falls down. When it spots a threat, it issues a social security number and sends you a notification. The whole experience is built to look exactly like the show. It is a fascinating look at computer vision running locally. No data leaves your machine. It is a cool way to see how this fictional tech actually works. Give it a try if you are into tech and privacy.

🆔 @hackernewsgithubprojects

📦 peach-mitao/peach

Peach: Your Private Media HQ

Peach is the local-first media system that puts your digital life back in your hands. It runs entirely on your own hardware, acting as a private media hub that automatically scans your hard drive to find and organize your files. The coolest part is its metadata engine, which quietly pulls in details, posters, and performer info from multiple online sources to build a rich, searchable library for you. You get a polished interface to watch and manage everything without your data ever leaving your house. It is the perfect setup for anyone who wants a Netflix-like experience with full control over their privacy.

🆔 @hackernewsgithubprojects

📦 oosexgang/hackios-xnu-userland

iOS 16.5 on an Android Phone?

HackiOS XNU Userland is the custom operating system layer that runs the real iOS sixteen point five interface on standard Android hardware. While the companion kernel repository handles the low level plumbing, this project builds the actual root and system images that boot the unmodified mobile software. It includes a custom initialization program, injected libraries for graphics and audio, and a package manager to make the experience feel authentic. The real surprise is the graphics bridge, which lets the phone render iOS animations using standard open source drivers instead of proprietary Apple code.

🆔 @hackernewsgithubprojects

📦 nousresearch/hermes-labs-plugins

Hermes Labs Plugins: Show It Once, Automate Forever

Show a browser workflow just once and let an AI agent learn it by watching, not by reading instructions. Hermes Labs Plugins includes a tool that records your mouse clicks and keystrokes in real time, safely masking any passwords you type. It then converts that visual demonstration into a repeatable skill, meaning you never have to describe a repetitive task in words again. The repository also features an inbox plugin that gathers scattered background jobs into one clear dashboard, and a secretary mode that drafts customer replies for your approval before sending. These are experimental tools built by Nous Research, offered to the community for adoption rather than official support.

🆔 @hackernewsgithubprojects

📦 atbfun/atbgameport

ATBGamePort: Stop Guessing, Start Porting

ATBGamePort is the game porting toolkit that actually makes old desktop games work in your browser without the usual glitches. Instead of letting an AI guess how your game logic works, which usually results in broken physics or black screens, this project forces a strict engineering process. It digs into the original executable file to understand exactly how the game runs, then rebuilds it with proper sound, saving, and mobile support. It even checks its own work to make sure nothing is broken before you play. It’s the difference between a messy toy and a real, playable web port.

🆔 @hackernewsgithubprojects

📦 enki013/hyperosunfcker

HyperOS Unfucker: Unlock Hidden Xiaomi Settings

This Android app bypasses manufacturer lock-ins to force your budget Xiaomi phone to run like a flagship, all without rooting your device. It uses the Shizuku tool to execute hidden system commands that unlock features like high frame rates, premium visual effects, and smoother multitasking that Xiaomi usually reserves for expensive models. You can also remove preloaded junk apps with a single tap. It’s a fascinating peek into the hidden controls of your phone, turning a standard device into a much faster machine. Just be careful, though, because pushing those visual settings too hard can make your phone run hot.

📰 https://news.ycombinator.com/item?id=49309357

🆔 @hackernewsgithubprojects

📦 faiface/ai-pair

AI Pair: Watch Your AI Type at Human Speed

AI Pair is the VS Code extension that turns AI coding assistance into a live, watchable collaboration. Instead of just handing off a task and reviewing a giant block of finished code later, this tool places a second, visible cursor in your editor. The AI agent types character by character at a comfortable, human speed, while explaining its reasoning out loud in a side panel. You can pause it, interrupt it, or even take over the keyboard to type yourself while the agent watches. This is built for people who want to deeply understand the code they are working on, rather than just accepting it blindly.

🆔 @hackernewsgithubprojects

📦 rlaope/deepseek-bot

deepseek-bot: The AI Teammate That Actually Remembers

A self-hosted AI teammate that lives inside Slack, Discord, and Telegram, and actually remembers what you discussed last week without you repeating yourself. DeepSeek bot plugs directly into the chats you already use, so you never have to open a new tab to ask for help. The coolest part is that it keeps a persistent memory of your conversations, meaning it can recall specific decisions from past threads and apply them to new questions instantly. It even drafts real documents like spreadsheets and slides right inside the chat.

🆔 @hackernewsgithubprojects

📦 datologyai/zephon

Zephon: The Data Loader That Doesn't Lie

Zephon keeps your training runs honest when your hardware changes mid-job. Standard data loaders often shuffle the deck or crash when you resume a training run on a different number of GPUs, which ruins your ability to compare experiments fairly. Zephon solves this by guaranteeing that every batch of data is identical no matter how many machines you throw at it. It handles tokenization and data mixing on the fly, so you skip the expensive pre-processing step entirely. If your cluster gets interrupted and you restart on fewer cards, it picks up exactly where it left off with the same data.

🆔 @hackernewsgithubprojects

📦 steventsvik/ghosttyextreme

Ghostty Extreme

Ghostty Extreme is the macOS terminal that lets you watch AI coding agents work live in real time. Built as a specialized version of the fast, native Ghostty terminal, this app connects directly to tools like Claude Code and Codex to show you exactly what your agents are doing across multiple projects. As an agent works, a dedicated sidebar shows its thinking process, while a built-in code editor physically types out the agent's edits as they happen. It even maps out which files are being read, tracks your backend services, and lets you approve or deny permissions with a single click.

🆔 @hackernewsgithubprojects

📦 tencentcloud/octop-harness

Octop Harness: The Missing Layer for Real AI Agents

Tencent Cloud released Octop Harness, a production-grade runtime that finally turns experimental AI agents into reliable, working software. It wraps the deepagents foundation in a layer of real-world engineering, handling memory, security, and multi-provider model routing so your agents actually run without breaking. You get a full terminal interface, pluggable storage backends, and a registry to manage multiple isolated agents in one go. It solves the classic problem: demo agents work, but production agents fall apart. This is the glue that makes them stick. If you are building agents, this is the safety net you have been waiting for. Stop building sandboxes and start building products. This is the one.

🆔 @hackernewsgithubprojects

📦 startluxlabs/decisionpivot

DecisionPivot: The Sneaky Test for AI Logic

DecisionPivot exposes when artificial intelligence fakes its reasoning. Instead of asking open-ended questions, this diagnostic benchmark feeds models pairs of scenarios that differ by just one tiny detail. If a model gets both versions right, it truly understood the situation. If it fails, it likely guessed based on surface patterns. This catches shortcuts that standard tests miss. It proves whether a system can handle subtle changes in real-world data. It’s the difference between a smart assistant and a parrot. Use this to check if your AI actually thinks or just vibes. Stop relying on basic benchmarks.

🆔 @hackernewsgithubprojects

📦 zeamoxwang/pdmd

Generate Videos in Two Steps with PDMD

Generate a complete video with sound in just two network evaluations. That is the core promise of PDMD, a new training method for video diffusion models. Standard fast generation often leads to blurry, oversaturated images because the learning process drifts over time. PDMD solves this by filtering out the specific error component that causes that degradation. It is essentially a one-line fix to the training loop that makes the student model learn much more stably. The result is a video generator that is both incredibly fast and surprisingly high quality. You can even run it on a single, modest graphics card.

🆔 @hackernewsgithubprojects