📦 harnessrouter/harnessrouter

HarnessRouter: Run AI Agents Locally

HarnessRouter puts professional AI coding agents directly onto your own hardware, letting you run tools like Codex and Claude Code without any cloud account. It works by spinning up a single Docker container that gives these agents a real workspace with terminal access and file systems, not just a chat window. You plug in your existing API keys, paste your prompts, and watch the agent write code, run commands, and debug errors live in a browser console. The project stands out because it implements an open standard for how these agents talk, meaning you are not locked into one vendor’s proprietary system.

📰 https://news.ycombinator.com/item?id=49335595

🆔 @hackernewsgithubprojects

📦 kevinywu/manosim

Simulating Real Human Hands with Physics

Human hands are soft, yet physics engines typically treat them as stiff, brittle objects that shatter on impact. Manosim solves this by converting detailed digital hand models into a format that rigid-body simulators can actually understand. The project takes a standard hand model and rebuilds it as a physical object with realistic weight, inertia, and smooth collision surfaces. This allows researchers to test how robots might interact with human hands using accurate physics, rather than simple geometric shapes. It bridges the gap between digital animation and physical reality. You get a realistic hand you can actually simulate, not just look at.

🆔 @hackernewsgithubprojects

📦 tiger3807861189/deepseek-v4-j-space-capability-realization-report

DeepSeek V4 J-Space Capability Realization Report

DeepSeek V4 is so powerful that its main problem is actually its own unpredictability. This new report from a developer named Tiger proves that a tool called J-Space can fix this by keeping the model focused without changing its core brain. Think of it like giving a brilliant but distracted expert a strict checklist and a memory aid. By organizing how the model thinks, checks its work, and remembers where it left off, the setup helps it finish complex coding and reasoning tasks without losing the plot. The results show it beats several top-tier competitors on tough benchmarks.

📰 https://news.ycombinator.com/item?id=49336380

🆔 @hackernewsgithubprojects

📦 lexmount/moli

Moli: The Browser That Saves AI Ten Megabytes

Moli is a headless browser built in Rust that runs on just seventy-three megabytes of memory while matching Chrome’s speed, solving a huge problem for AI agents. Instead of constantly rendering the visual world, it treats the page structure as the source of truth and only calculates visuals when you truly need a screenshot. This means your AI can crawl, search, and automate tasks in a fraction of the cost without the heavy overhead of a standard browser. It is a complete runtime that fits in your pocket, making it the perfect tool for anyone building automated web workflows that need to be fast and efficient.

🆔 @hackernewsgithubprojects

📦 ai-evals-course/evals-skills

How to Find AI Mistakes with Evals Skills

Evals Skills is the agent guide that finally turns raw AI outputs into actionable error insights. It stops you from guessing what went wrong and instead builds a custom review tool to help you find it. The standout feature is the error discovery mode, where you point an AI agent at your data and it designs a visual interface to help you spot mistakes. As you review samples and leave notes, the agent organizes your feedback into failure categories and suggests new items to review. This turns a chaotic manual process into a structured workflow that catches subtle issues you would likely miss alone.

🆔 @hackernewsgithubprojects

📦 outthislife/brooklyn-skills

Brooklyn's Skills: The Missing Instruction Set for Coding Agents

Brooklyn's Skills solves the problem of inconsistent AI coding output by providing portable instruction packages that force coding agents to follow strict engineering standards. Instead of constantly retraining your AI to clean up code or manage pull requests, you simply hand the repository to it. The standout feature is the babysit skill, which actively monitors your pull request until the build is green, handling failed tests and review comments automatically. This turns a chaotic development process into a disciplined workflow where the agent behaves like a senior engineer, ensuring nothing ships without proper cleanup and verification.

🆔 @hackernewsgithubprojects

📦 zouyuxuan122/deepseek-harness-eac

Deepseek Harness EAC: The All-In-One AI Desktop

Deepseek Harness EAC is the desktop app that turns a complex command-line AI tool into a simple double-click experience. It bundles the entire runtime, so you do not need to install anything else on your machine. Just click, and the app opens a native window for your AI agent. The coolest part is the built-in plugin marketplace, which lets you instantly install custom looks, memory tools, and even a live terminal right inside your chat. It is like giving your AI assistant a full workspace instead of just a text box. If you want to use this powerful AI without wrestling with code, this is your one-stop shop.

🆔 @hackernewsgithubprojects

📦 syv-ai/qwen38-27b-rtx3090

Qwen 38 27B RTX 3090

Running a massive twenty seven billion parameter AI model on a single consumer graphics card sounds impossible, but this setup proves otherwise. The project, called qwen38-27b-rtx3090, gets a Qwen 3.8 model humming on one RTX 3090, achieving speeds of over four hundred tokens per second when handling many requests at once. It solves a major bottleneck by using clever quantization tricks that free up enough memory to let the card actually do its job, rather than just sitting idle. You get a standard API interface that works with existing tools, allowing you to run complex pipelines or chat locally without needing expensive enterprise hardware.

🆔 @hackernewsgithubprojects

📦 radishxn/run-abroad

Run Abroad: Find Your Visa in 10 Seconds

Filter seventy-one international immigration pathways based on ten simple personal details. Run Abroad takes your age, education, language skills, and savings to instantly show which visas you qualify for, and exactly what you are missing to bridge the gap. This browser-based tool removes the guesswork from navigating complex global residency rules, saving you from hiring expensive advisors for basic fact-checking. You get a clear, honest list of realistic options, sorted by your specific goals and budget. It is the fastest way to map your next move abroad without the noise.

🆔 @hackernewsgithubprojects

📦 tiger3807861189/j-space-cognition-suite-v3.6

J-Space Cognition Suite: The AI Control Layer

The J-Space Cognition Suite is the cognitive control layer that stops AI from losing the plot on long tasks. It’s not a new model, but a set of text rules that manages what an AI keeps in mind, making sure it doesn’t forget your requirements halfway through a project. The coolest part is how it handles state drift, where the AI forgets its own goals after a few steps. By externalizing its thoughts into a simple ledger, it remembers exactly where it is and what it has already verified, so it can pick up right where it left off.

🆔 @hackernewsgithubprojects

📦 pg83/solo

Solo: The Static Linux Binary Trick

Ship one static binary and let it use your existing GPU driver, completely bypassing the need for containers or duplicate system libraries. This project called Solo solves a massive headache for Linux developers by allowing fully static applications to load native graphics drivers at runtime without dragging in a second C library. It works by building a bridge that translates standard system calls into your application's base runtime, meaning you get hardware acceleration without the bloated packaging overhead usually required. The result is a single, clean executable that talks directly to your machine's graphics card, making deployment surprisingly simple.

📰 https://news.ycombinator.com/item?id=49337618

🆔 @hackernewsgithubprojects

📦 zju-real/hugagentos

HugAgentOS: Agents That Learn Their Own Rules

HugAgentOS is an agent system that builds its own rulebook as it works. Instead of relying on a fixed set of instructions, it maps out the specific relationships between concepts in a domain and uses that map to reason. When it encounters a new problem, it updates its own framework to ensure future answers are grounded in verified facts rather than guesswork. This self-evolving approach turns a standard chatbot into a tool that actually learns the structure of your specific field. The result is an assistant that gets more reliable over time, making it a fascinating look at how we can build AI that truly understands context.

🆔 @hackernewsgithubprojects

📦 gkartist75/wan2gp-desktop

Wan2GP Desktop Launcher: Run Local AI Video Generation in One Click

Run a powerful local AI video generator with a single click, thanks to the Wan2GP Desktop Launcher. This Windows tool eliminates the complex manual setup required for Wan2GP by automatically handling Git, Python, and CUDA installations for you. It detects your specific graphics card to ensure the right hardware drivers are installed, then opens a clean window where you can launch the application instantly. The real value lies in its automated environment management, which saves developers hours of troubleshooting. Instead of fighting with command lines and version conflicts, this launcher creates a ready-to-use workspace that updates itself seamlessly, making local AI generation accessible and straightforward.

🆔 @hackernewsgithubprojects

📦 truefoundry/trueforge

TrueForge: Stop Building Agent Plumbing

Building an AI agent usually means drowning in code just to manage memory and tool calls. TrueForge changes that by acting as a complete runtime layer that handles the heavy lifting for you. It manages model calls, sandboxes, and session states, so your agent actually works instead of just chatting. The coolest part is its sandbox-as-a-tool feature, which only spins up isolated execution environments when the agent truly needs to run code. This keeps your setup fast and your secrets safe, while letting you focus on the logic, not the plumbing. If you want to build real agents without building a house every time, this is your shortcut.

🆔 @hackernewsgithubprojects

📦 yetone/cumora

Cumora: The Chat App Where AI Agents Are Coworkers

Cumora is a team chat app where AI agents are actual employees, not just chatbots. You invite them to a group, and they start working alongside humans, grabbing tasks from a shared board, sending real emails, and chatting in the same threads. The coolest part is that these agents have persistent memory and can coordinate with each other without stepping on toes, whether they run on cloud servers or your own local machine. It is basically a digital office where the robots handle the busywork so you can focus on the big picture. If you want a team that never sleeps and never argues, this is how you build it.

📰 https://news.ycombinator.com/item?id=49338707

🆔 @hackernewsgithubprojects

📦 coreyhaines31/makerskills

Makerskills: Your AI Operator Brain

Give your AI agent the exact playbooks to handle your business decisions, research, and creative workflows. Makerskills is a collection of twenty structured skills that turns tools like Claude Code into a serious personal operator. It stops the agent from guessing and forces it to follow proven frameworks for everything from picking domain names to modeling financial scenarios. The coolest part is the Maker Council skill, which simulates a board of advisors like Paul Graham and Jeff Bezos to pressure-test your ideas. Instead of getting generic advice, you get specific, conflicting perspectives that force you to make a better call.

🆔 @hackernewsgithubprojects

📦 electricitysheep/dsh-handbook

The Missing Manual for DeepSeek's New AI Engine

DeepSeek just open-sourced a new engine that turns any AI model into a fully customizable, code-running assistant. The dsh handbook is the missing guide that teaches you how to actually use it. It’s not just a list of commands; it’s a deep dive into the framework’s inner workings, showing you how to build your own plugins, tweak performance, and run complex real-world tasks. The coolest part? You can see exactly how to cut costs by nearly half just by adjusting a few settings, and learn how to build your own tools from scratch.

🆔 @hackernewsgithubprojects

📦 dfin/neural-network-visualisation

Neural Network Visualisation: Draw a Digit, Watch the AI Think

Neural Network Visualisation is the browser tool that makes you actually see how an AI reads your handwriting. You draw a number on a simple grid, and instantly, a 3D model of a neural network lights up. It’s not just a black box guessing; you watch the raw pixels travel through the network, with colors showing how strongly each part is firing. It’s like watching a thought happen in real time. The coolest part? You can scrub a timeline to see the network learning. It’s a fun, tactile way to understand the magic behind digit recognition without reading a single line of code. Just draw, watch, and learn.

🆔 @hackernewsgithubprojects

📦 robonuggets/gauntlet-loop

The AI Trick That Refuses Good Enough

Gauntlet Loop is the prompt generator that forces your AI agent to stop settling for good enough output. The core problem with most AI tools is that they grade themselves against vague ideas, which leads to drift and weak results. This project fixes that by turning any goal into a strict competition. You pick a real, existing reference, like a specific website or article, and the system creates a prompt where a builder works on the task while a separate, harsh critic blind compares the new work against that reference. The loop keeps running until your version beats the reference, not just until the AI feels satisfied.

🆔 @hackernewsgithubprojects

📦 awesome-dsh-plugin/dsh-find-plugin

Dsh Find Plugin: The Agent Tool That Finds Plugins For You

Dsh find plugin is the discovery tool that lets your coding agent find and install new capabilities for you. You simply tell the agent what you need, such as a way to get notified when a task finishes, and it automatically searches the public GitHub ecosystem. The results are ranked by stars, so you see the most popular options first. Each result includes a short description and a ready-to-run install command that the agent can execute for you. If a plugin is on the curated awesome list, you even get a hand-written, bilingual description. It is a simple, live search that turns your agent into a plugin finder.

🆔 @hackernewsgithubprojects