📦 aaddrick/building-with-typesafe-jev

Teach Your AI Agent To Trust

Building with TypeSafe Jev stops your coding agent from treating a fast decision model like a slow chatbot. It teaches the agent to ask structured questions and get instant, confident answers instead of vague paragraphs. The coolest part is the prior art library, which sorts one hundred and fifty real projects by how they work, so a drone and a game bot share the same lesson. It turns guesswork into solid engineering. You get a smarter agent that builds reliable tools. Give it a try.

🆔 @hackernewsgithubprojects

📦 capturegrubenchant/solidworks

SolidWorks: Your AI Draftsman

Talk to your 3D modeler in plain English and watch it build real parts. This project plugs an AI assistant directly into a live SolidWorks instance, letting you command complex engineering tasks like sketching shapes or extruding bosses without writing a single line of code. The magic lies in how it handles the messy details. SolidWorks commands often require dozens of parameters that crash standard connections. This server intelligently routes simple commands directly but automatically generates temporary scripts for the heavy lifting, ensuring your AI never gets stuck. It turns a rigid, menu-driven software into a responsive collaborator.

🆔 @hackernewsgithubprojects

📦 hasanyilmaz/operon

Operon: The Task Manager Built For AI Agents

Operon is the Obsidian plugin that finally makes your task list speak the same language as your AI agents. Most to-do apps are built for humans clicking buttons, but Operon flips the script by giving software a safe, structured way to read and update your notes without breaking anything. This means you can actually let an AI assistant help you organize your day by moving tasks around or checking them off, all while keeping your notes as plain text you own forever. It’s the rare tool that treats your digital assistant as a teammate rather than a threat.

🆔 @hackernewsgithubprojects

📦 justlovemaki/prismflowagent

PrismFlowAgent: Your Automated News Concierge

Transforms scattered web noise into polished, publishable articles. PrismFlowAgent watches GitHub trends, RSS feeds, and social media, then uses large language models to summarize the most relevant stories into structured content. It handles the heavy lifting of data ingestion, AI analysis, and formatting, allowing you to focus on final approval rather than manual research. The standout feature is its ability to push these finished pieces directly to WeChat, GitHub, or RSS feeds automatically. This turns a chaotic daily reading habit into a streamlined content production workflow. If you want high-quality updates without the manual grind, this is your answer.

🆔 @hackernewsgithubprojects

📦 bro3886/ical

Why ical is the fastest way to manage your Mac calendar

There is a command line tool called ical that makes your Mac calendar thirty times faster than the standard scripting method. It is a native tool that talks directly to your Mac’s calendar engine, skipping the slow middleman that usually handles these requests. You can create, update, or delete events using simple, natural language like next Tuesday or in two hours. It even lets you join meetings, check who is free, and invite people to events all from your terminal. It is perfect if you want to script your schedule or just want a faster way to manage your day.

🆔 @hackernewsgithubprojects

📦 pathwaycom/arc-task-gen

arc task gen

Create fresh, original logic puzzles that perfectly mirror the public ARC benchmark. The problem is simple: public test sets get scraped into training data, so high scores might just mean a model memorized the answers. This tool solves that by generating brand new tasks with the exact same distribution of colors, grid sizes, and complexity as the real set. It even uses semantic checks to ensure every new puzzle is genuinely unique and not a copy. You get a private evaluation set where a model cannot have seen the questions before.

🆔 @hackernewsgithubprojects

📦 firelex/jeff

Jeff: The Tiny AI Classifier That Decides in Milliseconds

Jeff is the ultra-fast decision engine that turns plain English into instant, confident choices without generating any text. You simply describe a situation and list your options, and Jeff returns a calibrated probability for each one in a single pass. This is what makes it perfect for routing support tickets, moderating content, or handling voice commands where speed is everything. On a standard workstation GPU, it answers in about twenty-two milliseconds, making it incredibly fast and easy to slot directly into your own code. Because it relies on a single forward pass rather than complex reasoning, it handles classification tasks with surprising accuracy and zero parsing overhead.

📰 https://news.ycombinator.com/item?id=49883844

🆔 @hackernewsgithubprojects

📦 michaelkinsy/pig

Pi: The Go Port of the Pi Coding Agent

Pi is a complete Go implementation of the Pi coding agent, originally built in TypeScript. It is not a rough translation or a partial rewrite, but a strict port where the original behavior is the contract. The goal is to run the same agent logic in Go, preserving every detail of how the original works. This matters because it proves the agent is language agnostic and allows developers to use a single, reliable backend across different systems. It is a clear example of faithful code migration, showing that complex tools can move between languages without losing their core identity.

🆔 @hackernewsgithubprojects

📦 qwdwqfwq/topconf-paper-figure-gallery

Top-Conf Figure Gallery: See How Top Researchers Design Figures

Three thousand five hundred sixteen hand-selected title figures from the biggest artificial intelligence conferences in the world are now available in one searchable place. The Top-Conf Figure Gallery is not just a folder of images, but a study guide for scientific communication. It filters out messy default charts to show you the clean, professional diagrams that win awards. You can search by conference or year to find exactly how top teams visualize their complex ideas. This is the perfect reference for any researcher who wants their paper to look as sharp as the industry leaders.

🆔 @hackernewsgithubprojects

📦 ng-native/ng-native

Angular Native: Real Native Views, No React

Angular components can now render as actual native views on phones without touching React at all. This project, called Angular Native, hooks directly into the underlying rendering engine so that your code builds real operating system elements instead of simulated web pages. You write standard Angular components and styles, but they compile into genuine mobile interface parts. The result is an app that feels and performs like it was written specifically for the device. It solves the problem of web frameworks feeling out of place on mobile. If you are building cross-platform apps, this is the bridge you have been waiting for.

🆔 @hackernewsgithubprojects

📦 grmchn/dopa-drill

Dopa Drill: The Math Game That Gets Louder

Solving math problems usually feels static, but Dopa Drill turns every correct answer into a loud, visual explosion. This browser-based tool acts like a math drill where the screen gets more chaotic and the sound gets louder the more you get right. It is built for elementary school students who need to practice their arithmetic, offering a unique twist where the game itself amplifies based on your performance. Instead of a quiet worksheet, you get a party of effects and audio that grows with your success. This makes practicing basic math feel like a game show rather than a chore.

🆔 @hackernewsgithubprojects

📦 backnotprop/bro

Bro: The Anti-Confusion Plugin for AI

Bro is the AI companion that finally keeps you and the code in sync. When your AI assistant starts speaking in jargon or drifting off-topic, Bro steps in to translate. It forces the AI to repeat your request in plain English, ensuring you both agree on the goal before a single line of code is written. It even spins up a fresh, unbiased agent to check decisions, stripping away any bias from the current chat. It’s like hiring a translator who actually understands the project. If you are tired of AI hallucinations or confusing outputs, this is your sanity check. It turns chaotic coding sessions into clear, manageable steps.

📰 https://news.ycombinator.com/item?id=47870583

🆔 @hackernewsgithubprojects

📦 uw-syfi/vibesys

Can AI Agents Build Bespoke Systems? Vibesys

Vibesys lets AI agents build custom software systems by orchestrating specialized coding and profiling tools to iteratively optimize performance. This project solves the problem of creating tailored, high-performance applications by allowing developers to define unique skills and agents that work together in a sandboxed environment. It is useful because it moves beyond generic prompts, enabling precise control over how AI agents handle specific technical tasks like microservice optimization or serving systems. The most interesting part is its ability to evolve code through hypothesis testing and real-time profiling, essentially teaching the agents to improve their own work.

📰 https://news.ycombinator.com/item?id=48096598

🆔 @hackernewsgithubprojects

📦 aitofound/scienceide

ScienceIDE: The Virtual Lab for AI

ScienceIDE turns the world's most complex scientific code into a safe playground where artificial intelligence can practice doing real research. Instead of just reading about chemistry or physics, these agents run actual simulations in a sandbox, learning from their mistakes without breaking anything. This project solves the problem that AI models struggle to apply theoretical knowledge to practical scientific tasks. By creating these learnable environments, we can build AIs that genuinely understand how to run experiments and analyze data, not just talk about science. It’s a big step toward having a digital lab partner that can handle the heavy lifting of discovery.

🆔 @hackernewsgithubprojects

📦 hokindeng/object-permanence

Object Permanence in World Models

Object permanence is the computer vision framework that finally teaches artificial intelligence that physical things still exist even when they are hidden from view. While typical video models easily lose track of objects the second they slide behind an obstacle, this clever project provides a complete physical data factory using over one hundred fifty specialized Blender simulations to generate complex, occluded scenarios. By training a lightweight video model on this dataset using AWS Trainium chips, it teaches the AI to accurately predict where hidden objects are and when they will reappear. It is a brilliant, open-source leap toward giving AI actual common-sense intuition about the physical world.

🆔 @hackernewsgithubprojects

📦 dzhng/jevgrep

Jevgrep: Find Code by Asking

Jevgrep is the command line tool that lets you find code by simply asking what it does. Instead of guessing file names or hunting through folders, you just describe the behavior you need in plain English. It then scans your project and hands back the exact files and code snippets responsible. This is built specifically to help coding agents skip the expensive guessing phase and jump straight to writing the fix. If you are using AI to write code, this tool saves you time and money by giving your agent the right context immediately. It is a clever, practical way to stop searching and start building.

🆔 @hackernewsgithubprojects

📦 rodmena-limited/stabilize

Stabilize

Build rock-solid automated pipelines that never break when orchestrating complex AI agents. This Python tool gives you an incredibly lightweight workflow engine that coordinates tasks using structured diagrams, making sure everything runs in the exact order you need without heavy overhead. While standard tools are either too simple or way too bloated, this one hits the perfect sweet spot with transactional safety, built-in circuit breakers, and automatic retries. You can even pause for human approvals, stream live data, or replay events like a time machine to squash bugs. It is the ultimate reliable backbone for your next smart agent system.

🆔 @hackernewsgithubprojects

📦 angusdevgo/seep-tool

Seep Tool: The Ultimate Client-Side Security Auditor

Seep Tool is the Windows security workbench that turns client-side enforcement flaws into a solvable puzzle. It targets the specific vulnerability where apps rely on their own code to verify licenses, bypassing proper server-side checks. The tool automates the entire process, from detecting installed software to applying precise patches that unlock premium features without touching the original binaries. It even handles the tricky part: a one-click restore function that returns your system to a pristine state, wiping all traces of the modification. This makes it a fascinating sandbox for developers to understand how these local authorization systems actually break down.

🆔 @hackernewsgithubprojects

📦 blacklabelhq/symphonyrecomp

Castlevania on PC? SymphonyRecomp

Run the classic PlayStation hit Castlevania: Symphony of the Night directly on your modern computer with SymphonyRecomp. This open beta project recompiles the original game code to work with current hardware, fixing the awkward limitations of older consoles. The standout feature is the new extended screen support, which lets you see more of the castle without enemies or objects popping in and out at the edges. It is a human-made effort that brings the beloved title to modern machines without needing a virtual machine. It is still in beta, so expect a few bugs, but it is a huge step for fans wanting a smoother experience.

🆔 @hackernewsgithubprojects

📦 jiarui-0410/multimodal-product-discovery

Multimodal Product Discovery

A fashion search engine that understands the difference between a red shirt and a red dress is a surprisingly difficult engineering problem, and Multimodal Product Discovery tackles it by learning from how real people behave. Instead of just matching keywords, this system builds a unique profile for every user by watching what they click, like, or buy, then combines that personal taste with visual understanding to re-rank results. It proves that adding this layer of personal preference dramatically improves the accuracy of what shows up at the top of your feed.

🆔 @hackernewsgithubprojects