📦 gkurt/tegaki

Tegaki

Turn any font into animated, self-drawing handwriting instantly without manually drawing a single line. This clever project processes font files behind the scenes by extracting glyph paths, converting them into clean digital skeletons, and automatically figuring out the natural stroke order and width. It means you can drop in your favorite font and get a beautiful, realistic writing animation that draws itself stroke by stroke. It supports all major web frameworks, works right in your terminal to output lightweight SVG files, and even features an interactive studio to customize your text. Check it out to bring your web projects to life.

🆔 @hackernewsgithubprojects

📦 yeje-cpu/talk-to-fengge

Talk to Fengge

Clone anyone's voice and personality to have highly realistic, real-time voice conversations with less than a second of engineering delay. By combining LiveKit for audio streaming, open-source voice cloning tools, and deep personality injection, this project lets you simulate complete characters instead of just generating robotic text-to-speech. You can feed it brief audio samples alongside speech patterns and catchphrases to turn an AI into a specific person you can actually talk to over a simulated phone call. Experience the future of natural, voice-cloned AI companions today by trying out the pre-built templates.

🆔 @hackernewsgithubprojects

📦 aiksa2090/agentic-swarm

Agentic Swarm

Imagine running five specialized AI agents that collaborate in an infinite brainstorming loop, entirely offline on your own computer. The project agentic-swarm lets you do exactly that by running Google Gemma-4 models locally. It solves the massive video memory problem of running multiple AI models at once by using a clever trick: it dynamically loads, runs, and completely clears each model from your graphics card before the next agent takes its turn. You get a beautiful dark-mode interface styled like WhatsApp with full real-time streaming, allowing you to inject ideas and steer the conversation. It is a brilliant way to brainstorm complex ideas locally without sharing your data with cloud...

🆔 @hackernewsgithubprojects

📦 androidmalware/openclaw_termux

openclaw_termux

Turn your Android phone into an autonomous AI assistant that you can control entirely through everyday chat apps like WhatsApp, Telegram, or Discord. This repository provides a simple, one-command script that installs a full Ubuntu environment inside Termux without requiring root access, setting up the OpenClaw gateway in just a few minutes. Once active, you can interact with your mobile device remotely by sending text messages to analyze files, trigger system commands, read incoming notifications, and capture photos directly from your phone. It is a brilliant way to recycle an old device into a private, always-on pocket server that acts on your commands.

🆔 @hackernewsgithubprojects

📦 zjunlp/lightmem2

LightMem2

LightMem2 is the lightweight runtime framework that finally stops context bloat from making your AI agents incredibly slow and expensive. When you run AI agents for a long time, their chat history and massive tool outputs quickly clog up the system, sending your token costs through the roof. This clever plug-and-play plugin acts like a smart filter, stabilizing your prompts and trimming away useless, oversized details before they can poison future turns. In real-world sessions, it slashes input tokens by over ninety-five percent and cuts hosting costs by eighty-seven percent. If you want faster, cheaper agents that do not lose their minds over time, this is the tool to install.

🆔 @hackernewsgithubprojects

📦 everyinc/hands-on-deck

hands-on-deck

hands-on-deck is the PowerPoint tool that lets AI agents edit and build slides using predictable JSON patches instead of fragile XML code. PowerPoint files are basically zipped XML, which makes it incredibly easy for an AI to corrupt a slide, miss brand guidelines, or mess up layout math. This tool solves that by letting agents write a simple, structured list of changes that run through a Python command-line interface. If an agent tries to modify a shape that does not exist, the tool rejects the entire patch, lists the actual shapes on the slide, and teaches the agent how to correct its own mistake instantly. It even includes an automatic...

🆔 @hackernewsgithubprojects

📦 anthropics/claudes-c-compiler

claudes-c-compiler

An artificial intelligence actually built a fully functioning C compiler entirely from scratch in Rust. The project, called claudes-c-compiler, was written completely by Claude Opus 4.6 with zero compiler-specific dependencies. It handles everything on its own—from the initial code parsing and optimization to assembling and linking final ELF executables for x86, ARM, and RISC-V. It is so complete that it can successfully compile a booting Linux kernel and build massive, complex codebases like PostgreSQL and SQLite. It is a mind-blowing demonstration of what machine learning can produce when pushed to the absolute limit.

🆔 @hackernewsgithubprojects

📦 mihaiciprianchezan/agentic-first-enterprises

Agentic First Enterprises

Build a company where AI agents run the work at machine speed while keeping humans in absolute control. This repository outlines agentic-first-enterprises, a fascinating operating model that treats organizational roles as interchangeable contracts. In this system, any seat can be filled by an agent, a traditional program, or a human. The clever part is the universal handbrake: any authorized human can pause any active process, inspect exactly what happened, make a manual correction, and resume the flow right from that spot. It perfectly balances autonomous speed with human accountability, showing how future companies can run safely and efficiently.

🆔 @hackernewsgithubprojects

📦 alanlusun/general-keypoint-detection

General Keypoint Detection

Computer vision models usually break down when you ask them to track something they have never seen before, but this new system can pinpoint critical joint and skeletal landmarks on almost any object imaginable. The creators of general-keypoint-detection built a massive dataset by merging twenty-nine different public sources, giving their AI over one million real-world examples to learn from. Instead of training a separate, specialized model for humans, another for cars, and another for animals, you get one unified brain that understands how things connect. It makes teaching robots and cameras to understand physical movement incredibly simple.

🆔 @hackernewsgithubprojects

📦 intelchina-ai/bittern

BitTern

Shrink massive artificial intelligence models down so they require barely any computer memory, without ruining their smarts or costing a fortune to train. Usually, compressing an AI model to just one point five eight bits per parameter requires rebuilding the whole thing from scratch, which takes massive supercomputers. But this project lets you take already trained, full-sized models and convert them into highly efficient three-value systems on a budget. It packages a clever, award-winning technique that makes this super-tight compression simple and highly accurate, allowing everyday computers to run powerful open-source models incredibly fast.

🆔 @hackernewsgithubprojects

📦 peetzweg/opendisplay

OpenDisplay

Turn your spare iPhone or iPad into a fully functional second monitor for your Mac without paying a single cent. This clever open-source tool lets you bypass expensive subscriptions and restrictive system requirements by creating a true virtual display you can drag windows onto. Just connect over Wi-Fi or plug in a standard USB cable for lag-free performance, complete with native Retina sharpness and responsive touch gestures like tapping and two-finger scrolling. It is entirely self-hosted, keeping your data private. Grab the download, plug in your phone, and instantly double your screen real estate.

🆔 @hackernewsgithubprojects

📦 oceanflowlab/taro

Taro Video Temporal Grounding

Multi-modal language models can accurately pinpoint specific moments in a video by learning to spot when temporal sequences are intentionally shuffled. A new framework called taro achieves this by training models to think with time instead of relying on random exploration. It creates high-quality reasoning paths using dense captions, then tests the model's understanding with a temporal sensitivity reward that shuffles frames near the target event boundaries. If the model is actually paying attention to the video's timeline, its confidence drops when the order is disrupted. This teaches the model to anchor its reasoning to precise temporal evidence, delivering state-of-the-art zero-shot performance.

🆔 @hackernewsgithubprojects

📦 shpigford/knockoff

Knockoff

Amazon is flooded with trademark-squat pseudo-brands like SZHLUX or LATTOOK, which are just random strings registered purely to sell cheap commodity goods with zero warranty or company behind them. A clever free browser extension called knockoff fixes this by filtering these fake brands right out of your search results. It runs entirely locally in your browser, using a smart name-heuristic scanner to score weird gibberish patterns like all-caps consonant runs, alongside daily community-curated brand lists. You can set it to hide, dim, or label these junk listings instantly, giving you back an Amazon filled only with real, established companies.

🆔 @hackernewsgithubprojects

📦 owl-listener/designer-skills

designer-skills

Turn your favorite AI assistant into a professional design partner that actually knows how to craft great user experiences. With designer-skills, you can load nearly a hundred specialized design skills and thirty direct commands straight into tools like Claude Code or Gemini. Instead of giving you generic advice, this clever collection teaches AI agents the nuanced judgment of real product designers, letting you generate accessible color palettes, wireframe interactive flows, and audit complex layouts with a single prompt. It is the ultimate shortcut for developers who want to ship polished, highly usable interfaces without having to guess the design rules.

🆔 @hackernewsgithubprojects

📦 cybertron957/agent-teams

Agent Teams

You can now run a self-hosted team of AI agents that collaborate entirely on your own hardware twenty-four seven. The agent-teams project acts as a mission control server with a real-time dashboard where you can watch autonomous agents browse the web, write code, run terminal commands, and coordinate clicks in a headless browser. To solve the problem of agents getting stuck on logins or security prompts, it features an embedded browser takeover that lets you clear CAPTCHAs live from the dashboard, even on a headless server. It is a fantastic tool for developers wanting to run complex, multi-agent automations safely on their own terms.

🆔 @hackernewsgithubprojects

📦 robbyant/lingbot-vla-v2

lingbot-vla-v2

Robots are finally learning to share a single brain. A clever project called lingbot-vla-v2 maps twenty wildly different robot setups, from dual-arm setups to full humanoids, into a unified fifty-five-dimensional control system. By training on sixty thousand hours of physical and human video data, this system translates complex sensory inputs into precise physical actions. It uses specialized mixture-of-experts layers to let different robot designs share universal skills while retaining their unique movement patterns. This means a single foundation model can control a robotic arm, a dexterous hand, or a mobile base with remarkable precision. It is a massive step toward making general-purpose robots adaptable and easy to deploy in the...

🆔 @hackernewsgithubprojects

📦 ahepi/deepreason

DeepReason

Force your AI to argue with itself on the record to map out the truth. DeepReason is a clever harness that takes a hard, open-ended question and forces an LLM to generate a distribution of rival explanations. Instead of letting the model settle on a lazy, confident guess, the system makes every single candidate answer state exactly how it could be proven wrong. Weak theories get mechanically argued down, survivors face off in rigorous head-to-head trials, and every single step is committed to an immutable, replayable log. It is the perfect tool for stress-testing complex research questions where you need honest disagreement instead of polished AI vibes.

🆔 @hackernewsgithubprojects

📦 itsthelore/wayfinderrouter

Wayfinder Router

Wayfinderrouter lets you stop paying top-tier cloud pricing for simple artificial intelligence prompts by routing them entirely offline. Instead of calling a slow, expensive referee model to decide where your query should go, this clever tool analyzes the actual structure of your prompt, like word count, lists, and code blocks, right on your machine. It generates a sub-millisecond complexity score to instantly send simple tasks to your fast, free local model while saving the heavy lifting for the cloud. It is a brilliant, zero-latency way to slash your API bills without changing your existing code.

🆔 @hackernewsgithubprojects

📦 hristo2612/jinn

Jinn

Jinn is the orchestration layer that finally lets you run your favorite AI agents as a coordinated company of employees. Instead of replacing your existing command-line tools like Claude Code or Grok, it acts as a central hub, giving them a real organizational chart with departments and a reporting hierarchy. You design the roles in simple files, and a chief of staff agent automatically delegates tasks down the chain and reviews the work before it reaches you. The best part is that it drives these agents through your existing flat-rate subscriptions inside a real terminal, saving you from unpredictable API bills. This turns a messy pile of terminal tools into...

🆔 @hackernewsgithubprojects

📦 pouyahasanamreji/continuum

Continuum: Give Your AI Agents Long Term Memory

Your coding agents are amnesiacs, but continuum fixes that by giving them a shared, persistent memory that actually survives across chat sessions. Running locally on your machine, this clever tool acts as an MCP server with a built-in vector database. When one agent finishes a task, it registers the lesson it learned—like a specific library quirk or a guard ordering issue. When you spin up a brand-new agent later, it instantly searches that memory to avoid making the same mistakes. No more copy-pasting instructions; just let your agents learn from each other in real time.

🆔 @hackernewsgithubprojects