298 subscribers
14K videos
14.6K links
Download Telegram
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ alex-d-c/silicodraw

SiliCoDraw: The Local AI Figure Editor

SiliCoDraw is a local editor that lets you build scientific figures by dragging elements while your AI assistant writes the underlying code. Instead of asking a chatbot for an image and hoping for the best, you get a live canvas where the AI edits the file as you move things around. It runs entirely on your computer, so your data never leaves your machine. This is great for making pathway diagrams, Gantt charts, and chemical structures that actually make sense. The biggest win is that you stay in control. You are not just prompting; you are collaborating.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ orchardlinux/orchard

Turn an iPad into a Linux Computer

Put a full Linux operating system on your old iPad or iPhone. The Orchard project provides the exact kernel patches and hardware drivers needed to make Apple’s custom chips run standard Linux code. Instead of just looking at the device, you can actually boot a real operating system like a minimal shell or a full desktop environment right on the hardware. This is fascinating because it turns consumer electronics into general purpose computers, giving you complete control over your own data and software. It works by connecting the device to a computer, which handles the complex boot process for you.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ osyanne/citiesskylines2-osm-toolkit

Real Zoning Maps for Cities Skylines 2

A free browser tool displays the actual zoning maps of seventy-seven real cities from OpenStreetMap directly inside Cities Skylines 2. It solves the problem of guessing how a city is zoned by letting you see exactly which land use belongs on every block. You can then export that map as a single image file, which a game mod stretches perfectly over your playable area to align with the terrain. This helps builders replicate authentic urban layouts instead of random designs. The project is completely free, requires no installation, and runs entirely in your browser.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ shakewingo/tadpole

Tadpole: Building an LLM Agent From Scratch

Tadpole lets you build a capable language model agent entirely from scratch using only plain Python code, without relying on heavy external frameworks. The most fascinating aspect is how this project documents the step-by-step process of training a small local model to become a reliable helper. By showing the exact stages of development and measuring performance gains at each step, it turns abstract machine learning concepts into a tangible, working tool you can actually use. It even runs as a chat bot on your phone, proving that complex AI systems can be built by hand.

πŸ“° https://news.ycombinator.com/item?id=50315665

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ isas1/daily-team

Daily Team

Collaborate with a brand new four-person creative team every single day to design and build unique, single-file web tools. This automated project uses a language model to generate a fresh, daily crew of specialists, like a packaging designer, a calligrapher, or even guest characters from history and literature. The daily team pitches ideas, debates the design, and builds a fully functioning, mobile-friendly webpage under one hundred kilobytes. You get immediate, interactive utilities like custom workout sheets and pace calculators. It is a brilliant way to see how distinct viewpoints can shape everyday software, giving you a clever new tool to try out every morning.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ cortexapps/engrams

Engrams: The Self-Hosted AI Factory

Run your own private factory for AI coding agents with Engrams. This open source platform spins up each agent in its own isolated virtual machine, giving you full control over the infrastructure. The coolest part is how it handles time. When your agent stops working, the system takes a snapshot of the entire machine and saves it. The next time you need to pick up a task, it boots back up in milliseconds, exactly where it left off. Because you host it in your own cloud, your data never leaves your account. It is a powerful way to manage coding tasks without renting a third party service.

πŸ“° https://news.ycombinator.com/item?id=49923064

πŸ†” @hackernewsgithubprojects
Media is too big
VIEW IN TELEGRAM
πŸ“¦ faded-penguin021/tideo-auto-brightness

Tideo Auto Brightness: Know Exactly Why Your Screen Is Dim

Take control of your phone’s screen by replacing the mysterious, black-box algorithms that usually guess your brightness with clear, predictable math. This app, Tideo Auto Brightness, solves the headache caused by screens that flicker or dim too aggressively by using deterministic calculations you can actually see. Instead of wondering why your display suddenly changed, you get live graphs showing exactly how the app is adjusting to the light around you. It’s a glass box for your Android experience, letting you trust the process because you can verify the logic. If you want to stop guessing and start understanding your display, this is the tool to try.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ stevyhacker/lokalbot

LokalBot: Your Mac's Local Memory

LokalBot turns your Mac into a private assistant that actually remembers what you did today. Instead of sending your thoughts to a remote server, it runs a language model right on your machine to keep a running log of your workday. The coolest part is that it listens to your meetings and captures your dictation, then uses that context to help with autocomplete and quick recall later. It’s a solid, open-source tool for people who want their computer to be a bit more aware of their daily workflow without worrying about data leaving their desk.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ paintdotnet/paint.net-on-wine

Paint.net on Wine: Linux and Mac Ready

Paint.net on Wine is the experimental bridge that finally brings the popular photo editor to Linux and Mac without needing a dual boot. The repository provides the essential downloads and issue tracker for this specialized build, which uses a compatibility layer to run Windows software on other operating systems. The standout feature is a built-in expiration timer that forces users to download fresh updates every twelve weeks, ensuring everyone tests the most current fixes. It is strictly for early adopters who want to try the editor on non-Windows systems and help squash bugs by reporting issues.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ heiner/mills8a

Mills 8A: The 1947 Font Revived

Every letter in the Mills 8A typeface is an average of roughly nine hundred thousand physical impressions, traced directly from scanned pages of the American Mathematical Society Bulletin published between 1940 and 1948. This project does not redraw the classic Monotype Modern font; instead, it reconstructs the exact shapes, ink weights, and spacing measured from the hot metal type itself. The result is a digital font that preserves the subtle baseline wobble and random variations of letterpress printing, bringing the precise look of 1940s mathematical papers to modern LaTeX documents.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ jamesondh/chimera-vocal-synth

This Voice Synthesizer Actually Simulates Your Throat

CHIMERA is the browser synthesizer that actually simulates your throat instead of just playing audio files. It builds every sound from a live physics model of your lungs, vocal folds, and airway, meaning you can drag your tongue and jaw around a 3D head to see exactly how speech is formed. The coolest part is the Bestiary mode, which lets you morph human voices into animal calls by tweaking the anatomy in real time. It’s a fascinating look at the engineering behind human speech, and it turns a dry acoustic concept into a playable instrument you can actually see working.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ vxtzq/crowdgpt

CrowdGPT: Training AI on Your Own Hardware

CrowdGPT is the open-source AI project that trains large language models using your own hardware instead of massive data centers. It works by distributing the training workload across a network of volunteer computers, allowing anyone to contribute their device’s processing power to building the model. This decentralized approach removes the need for expensive server farms or specialized hardware, making the creation of powerful artificial intelligence accessible to anyone with a standard computer. By turning individual devices into a collective training resource, it offers a transparent and community-driven alternative to centralized cloud computing for developing new language models.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ pwa-today/pwa-check

PWA Check

PWA Check scans your web app to find hidden gaps that stop it from feeling like a real, installed application. It inspects the manifest, service worker, and viewport settings to pinpoint exactly what is missing for offline support or a polished user experience. If a critical piece is broken, the tool flags it as a failure, making it easy to drop into your automated testing pipeline. This gives developers a clear, objective answer on app readiness instead of relying on guesswork. The project turns complex progressive web app requirements into a simple health check, ensuring your app is truly installable and reliable for every user.

πŸ“° https://news.ycombinator.com/item?id=48836490

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ wallawalla47/infernix

Infernix: Running Qwen Locally

Infernix runs a massive language model entirely on a single consumer graphics card. It executes the entire inference process locally on your own hardware, meaning no data leaves your machine and no cloud connection is required. This gives you complete control over your privacy and processing speed without paying for external server costs. The project solves the problem of heavy compute demands by using efficient code to squeeze high performance from standard hardware. It is a practical way to experience advanced artificial intelligence on a desktop. If you want to build local tools, this is a strong foundation.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ open-net-worth-ai/open-net-worth

Open Net Worth: Your Self-Hosted AI Finance Manager

Open Net Worth lets you run a personal AI finance agent on your own computer, keeping your data private while it handles the heavy lifting. Instead of trusting a cloud service with your bank logs, you install this locally to track every dollar and investment. The real surprise is that it does not just list your spending; it actually executes trades for you. It connects to your accounts, learns your budget rules, and places real market orders when you approve them. You get the power of a sophisticated trading assistant without sending your financial secrets to a third-party server.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ diygod/pinky

Pinky: The AI Phone Remote

Pinky is the open-source tool that gives AI agents the keys to your actual phone, but only when you say so. It works on Mac, Windows, and Linux, letting programs like Claude Code tap, type, and swipe on your Android or iPhone from your computer screen. The coolest part? You stay in control. Pinky watches the agent work on a mirrored window, holding up risky moves like payments until you click allow. It even keeps your banking apps hidden by default. It is lightweight, free, and records every step so you can see exactly what the AI did.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ citypulso/pulsoicon

Pulsoicon: 9955 Icons With Their Own Moves

Nine thousand, nine hundred and fifty-five icons each perform a unique animation written specifically for that shape. Pulsoicon solves the boring problem of static icons by making a bell ring, a calendar lift its rings, and a trash can open its lid right when you hover over them. It is a drop-in replacement for lucide-react, meaning you just swap the import line in your project to get moving graphics without changing your code structure. Since every icon is its own module, your app only downloads the animations you actually use, keeping your bundle tiny. It works on React, Vue, Svelte, Android, and iOS, so your whole stack feels alive.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ lxe/phoneuse

PhoneUse: Control Your iPhone From Linux

PhoneUse lets a standard Linux computer see and operate an unmodified iPhone without any jailbreaks or cables. The phone streams its screen over Wi-Fi, while the computer acts as a Bluetooth mouse and keyboard to send taps and keystrokes back. This setup means you can control your phone with a normal mouse or let an AI agent do it for you via a command line. It is a fascinating proof of concept that uses only standard phone features. No apps are needed on the phone itself. This project shows how two common wireless protocols can be combined to create a powerful remote control tool.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ giovannipasq/chunky

Chunky: See Your AI Data Before It Breaks

Chunky is the visual workbench that lets you actually see how your documents break down before feeding them to an AI. When you build systems that let a machine read your files, the biggest mistake is usually made before the reading starts. If a table gets scrambled or a page header gets copied three times, the AI will guess wrong later. This tool turns that invisible process into a picture, letting you compare different ways of cutting up your text and watch exactly where the errors happen. It helps you fix the mess while it is still cheap, so your final answers are accurate.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ anomalyco/isocast

Isocast: The Mac Recorder That Splits Your Screen

Isocast changes how you record your Mac by saving every single source as its own separate file. While standard tools merge everything into one video, this app keeps your screen, camera, and windows distinct and perfectly synced. This means you can fix a bad take in your editor without re-recording the entire session. Written in Rust for Apple silicon, it runs on about half the CPU power of typical alternatives, leaving your computer free for other work. It also imports your existing setups automatically, so switching is painless. If you create content on a Mac, this tool saves time and simplifies your editing workflow.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ synchopate/llama-modchip

llama-modchip: Run 284B AI Models on Your CPU

llama-modchip is the CPU inference engine that squeezes massive AI models into standard computer memory without any graphics card. It shrinks a 284 billion parameter model so it fits entirely inside sixty-four gigabytes of RAM. This is interesting because you can run powerful, real-time AI on the laptop or desktop you already own. The software uses clever math to compress the model, letting it think fast without expensive hardware. You get local, private, and fast AI capabilities on plain hardware. It proves you don’t need a supercomputer to run big AI. Now you can run your own smart assistant without the cloud. That’s the power of llama-modchip.

πŸ“° https://news.ycombinator.com/item?id=50411961

πŸ†” @hackernewsgithubprojects