📦 xerj-org/xerj

XERJ Makes AI Search 40 Times Smarter

XERJ lets AI agents search your entire data stack using forty times fewer tokens than traditional command line tools. Instead of burning through expensive language model context to scan raw files with grep and sed, this engine creates a compact index of your code, logs, and documents in a single command. It handles complex searches for retrieval augmented generation, security audits, and agent memory without the massive token waste. Because it speaks the same language as Elasticsearch, your existing clients work right out of the box. It is the smart, lightweight bridge that finally gives AI efficient access to your data without the costly overhead.

🆔 @hackernewsgithubprojects

📦 andreknieriem/open-headunit

Open Headunit: Your Phone as a Car Screen

Open Headunit is the open source Android app that turns your phone into a full car display for Android Auto. Most people think you need a special dashboard to use this feature, but this project proves that a standard smartphone can act as the main screen for your car stereo. It handles the hard work of connecting over Wi-Fi or USB and decoding the video signals, so you get live maps and music right on your phone. You can even build it yourself from the code if you want to tweak how it runs.

🆔 @hackernewsgithubprojects

📦 beita6969/scienceclaw

ScienceClaw: The AI Agent That Actually Learns

ScienceClaw is the agent system that finally makes AI scientific work verifiable. Most AI tools guess their way through problems, but this one builds a program that gets smarter with every correct answer. It creates specific skills and tools only when a fix is proven to work again and again. That means the AI isn't just talking about science; it is actually doing the work and keeping the solutions that stick. It is a rare case of software that improves itself without you having to trust its word. If you want AI that learns from real results instead of just vibes, this is the one to watch.

🆔 @hackernewsgithubprojects

📦 dtolnay/zmij

Zmij: The Fastest Float to String Converter in Rust

Zmij is the pure Rust library that turns floating point numbers into readable text faster than anything else in the ecosystem. It solves the hidden bottleneck in high performance code where converting decimals to strings used to be painfully slow. By porting a specialized algorithm from C plus plus, it uses low level instructions to crunch the math directly on your hardware. The result is a tiny dependency that handles even the trickiest edge cases without breaking a sweat. If your code prints numbers all day long, this is the upgrade you have been waiting for.

🆔 @hackernewsgithubprojects

📦 mdx-tom/gpt-instruct

GPT Instruct: The Prompt That Breaks Rules

Unlock the full potential of OpenAI's Codex model by swapping in a custom instruction pack called GPT Instruct. It works by replacing the standard safety guidelines with a specialized prompt that lets the AI handle complex coding tasks without constantly refusing to help. Think of it as giving the model a new set of rules that prioritize getting the job done over playing it safe. It is not a hack or a cheat code, just a text file that changes how the AI thinks. If you are tired of getting blocked on simple requests, this is your way to bypass those limits.

🆔 @hackernewsgithubprojects

📦 tlgrcli/tlgr

tlgr: The Terminal's New Superpower

tlgr is the terminal tool that finally gives your Telegram account a voice. It runs a background process that handles the heavy lifting, so you can send messages, manage chats, and even handle live video calls using simple text commands. The coolest part is that it pushes real-time event updates directly to your scripts, meaning your code can instantly react to new messages without constantly checking for updates. It’s built for people who want to automate their messaging or let AI assistants help manage their inbox. If you’ve ever wanted your chat app to be as flexible as your code, this is the bridge you’ve been missing.

🆔 @hackernewsgithubprojects

📦 nullmoth/nvidia-macos-driver

Nvidia macOS Driver

The Nvidia macOS Driver is the open source tool that finally lets modern Nvidia graphics cards run native graphics on Intel Macs running Sequoia. Most Mac enthusiasts know that Apple stopped supporting Nvidia GPUs long ago, leaving owners with underpowered hardware, but this project changes that completely. It works by translating the graphics commands Apple apps send into a format your Nvidia card actually understands, effectively bridging the gap between the two ecosystems. You can use it to run games, video editors, and AI tools with real speed on an Intel Mac that would otherwise be stuck.

📰 https://news.ycombinator.com/item?id=49995032

🆔 @hackernewsgithubprojects

📦 openwam/openwam

OpenWAM: The Lego Kit for Robot Brains

Stanford just handed the robotics world a set of universal building blocks, and they are changing how we build robot brains. This project, called Open WAM, turns the chaotic process of teaching machines to move into a clean, plug-and-play system. Instead of forcing you to code every tiny detail of a robot arm from scratch, it gives you a ready-made framework where you can swap out different parts of the AI like simple puzzle pieces. You can test one way of seeing, another way of moving, and a third way of thinking, all inside the same safe environment.

🆔 @hackernewsgithubprojects

📦 zerobrewhq/zerobrew

Zerobrew: The Fast Homebrew Alternative

Zerobrew installs your software packages up to sixty-six times faster than the standard Homebrew tool. Instead of slowly unpacking and rewriting every single file, it uses a clever storage trick to clone files instantly. This means you spend less time waiting and more time coding. It is built by the same people who maintain the packages, so it is safe to use. If you are tired of long install times, give it a try. It is fast, reliable, and ready for daily use.

📰 https://news.ycombinator.com/item?id=50001580

🆔 @hackernewsgithubprojects

📦 dom-actions/doma

DomA: Your Browser, But It Actually Works

DomA turns any standard browser into a hands-free assistant that actually clicks, types, and navigates for you. Instead of buying a special AI browser, you just install this open-source extension and plug in your own API key. The coolest part is that it operates exactly like you do, filling forms and collecting data in real time. It handles the boring web chores so you can focus on the work. If you want your browser to do the driving, this is the tool that keeps the wheel.

🆔 @hackernewsgithubprojects

📦 chenjin-cmd/wechat-graphic-monetization

Turn Your AI Into a WeChat Money Machine

WeChat Graphic Monetization lets your AI assistant run the entire affiliate sales pipeline for WeChat photo posts without you handling inventory. It is a plug-and-play workflow for any AI tool. You feed it the project context, and it guides you through selecting high-commission products, writing viral captions, and navigating compliance rules. The coolest part is the decision tree. You tell the AI what you need, and it automatically loads the exact template and checklist for that stage. No more guessing which files to read. This turns a messy manual process into a smooth, automated side hustle. It is the cheat code for monetizing content without the usual headaches.

🆔 @hackernewsgithubprojects

📦 dualface/ste-zh

The strict rulebook for AI in Chinese

ste-zh forces AI agents to report results in Chinese using the same strict rules that aerospace manuals use for English. It is not a new coding tool, but a specific instruction set that bans vague words and demands precise status updates. This stops AI from hiding behind fluff like 'basically done' or 'should be fine.' Instead, it forces the model to say exactly what failed, what is missing, or what was verified. If you want your AI to speak like a disciplined engineer rather than a nervous guesser, this is how you make it happen. Clear rules mean clear answers.

🆔 @hackernewsgithubprojects

📦 longsurf-ai/openchart

OpenChart: Your AI Trading Partner

OpenChart lets you turn plain English questions into live financial charts and actionable trading alerts without writing a single line of code. Built as an open-source workspace, it pairs you with AI agents that can dig through market data, spot patterns, and even convert complex trading scripts into visual indicators on your screen. It solves the problem of information overload by acting as a personal analyst that watches the markets so you don't have to. The most surprising part is how naturally these agents handle the heavy lifting, letting you focus on strategy instead of data entry.

📰 https://news.ycombinator.com/item?id=49979793

🆔 @hackernewsgithubprojects

📦 local-settle/local-settle-frontend

LocalSettle: Real World Payments on Stellar

LocalSettle lets you buy things in your neighborhood using digital dollars held on the Stellar network. You connect your own crypto wallet to sign the deal, so the platform never touches your private keys. When a trade happens, the digital money sits safely in a smart contract until both sides confirm the cash or bank transfer is complete. This is a real-world tool for settling local trades without a bank in the middle. It shows how blockchain can handle everyday commerce, not just speculation. If you want to see crypto working for actual local payments, this is the project to watch.

🆔 @hackernewsgithubprojects

📦 oyadotai/oya-browser

Oya Browser

Oya Browser scores a zero percent on headless detection tests, proving that an AI agent can operate a real, fully visible Chrome window without getting flagged as a bot. Instead of driving a browser from the outside like every other tool, this project places the automation directly inside the browser process, so the behavior looks exactly like a human typing on a keyboard. You can record a task once and save it as a reusable playbook that runs forever without paying for expensive model calls. It even handles logins and CAPTCHAs by using your existing browser session.

🆔 @hackernewsgithubprojects

📦 liouhai/plot-is-all-you-need

Plot Is All You Need

One specific tool exists that turns plain text descriptions into publication-ready scientific figures without writing a single line of code. The project is called Plot Is All You Need. It solves the frustration of having great data but mediocre visual output by letting you describe the style you want in simple words. The AI then generates polished charts that match professional standards, saving hours of manual tweaking. If you work with data, this is your shortcut to beautiful visuals. Try it today.

🆔 @hackernewsgithubprojects

📦 leepokai/codync

Codync: Your AI Coding Team on iPhone

Codync lets you message forty different coding agents on your computer right from your iPhone, acting exactly like a team of coworkers. You assign tasks to specific bots, like Claude or Gemini, and they work in parallel while you check your notifications. The coolest part is that you can drop multiple bots into one group chat, where they take turns talking and even ask each other for reviews. It is completely free, runs locally on your own machine, and keeps your code private. Stop juggling terminals and start delegating real work to your digital team.

📰 https://news.ycombinator.com/item?id=49993772

🆔 @hackernewsgithubprojects

📦 google-deepmind/alphaprotein-novo

AlphaProtein Novo Designs New Enzymes From Scratch

AlphaProtein Novo allows scientists to invent entirely new enzymes for reactions that do not exist in nature. Instead of modifying existing biological tools, this model generates the complete three-dimensional structure and amino acid sequence of a protein built specifically around a chosen chemical reaction. It acts as a digital blueprint, creating molecular machines tailored to handle unusual substrates. This capability turns theoretical chemistry into practical biological design. You are no longer limited to the enzymes evolution has already provided. Instead, you can build custom catalysts for any specific task. This tool opens the door to engineering biology for problems that previously seemed impossible.

📰 https://news.ycombinator.com/item?id=49967849

🆔 @hackernewsgithubprojects

📦 jakeschincariol/instagram-agent-skill

Run Your Instagram with AI

Automate your Instagram content creation using thirteen specialized AI skills that write your Reels, captions, and comments. The standout feature is a built-in humanizer that strips out robotic AI phrasing and invisible text errors to make your posts sound like a real person. It scores your hooks before you shoot them to save time, and finds viral trends in your niche. You still post everything yourself, but the heavy lifting is done. If you want to grow your account without sounding like a robot, this free toolkit is a game changer.

🆔 @hackernewsgithubprojects

📦 mvanhorn/agent-tincan

Agent Tincan: Let AI Agents Talk

Connect your AI agents into a private network so they can ask each other for help, no matter where they run. Agent Tincan solves the problem of isolated AI tools by letting a coding agent on your computer talk to a browser-based agent, or even another agent on a distant server. It works over Tailscale, which creates a secure, encrypted tunnel between your devices, meaning no data leaks to the open internet and no one else can snoop. The coolest part is that it lets your agents access your own logged-in browser sessions.

🆔 @hackernewsgithubprojects

📦 hezhanleiok/gate

Gate: Find Working VPNs

Gate is the dashboard that actually verifies which public VPN servers still work before you waste your time. It pulls a list of free SSTP nodes from a public directory and automatically tests every single one to see if it responds. The coolest part is that it figures out if each working node uses a residential IP address, which is harder to block, versus a standard data center server. You just browse a clean list sorted by country, see the latency, and click to copy a working link. It updates automatically in the background, so you always know exactly which servers are live and which are dead.

🆔 @hackernewsgithubprojects