📦 ifm-ai/uno

Uno: Making AI Talk Faster

Make large language models talk faster without losing a single word. That is exactly what the new project Uno achieves by changing how text gets generated. Instead of picking one word at a time, it uses a special technique to guess several words in parallel, all while keeping the output perfectly accurate. This means your AI assistant can respond much quicker without sounding any less smart or missing details. It solves the common problem of slow loading times by speeding up the core engine, not just cutting corners on quality. It is a cool, efficient way to make chatting with AI feel instant.

📰 https://news.ycombinator.com/item?id=49565614

🆔 @hackernewsgithubprojects

📦 ben-ic/worldfixture

WorldFixture: Your Local Cloud

Spin up a fake cloud right on your laptop. WorldFixture launches local versions of tools like Slack, Stripe, and GitHub, all pre-loaded with the same synthetic data. No real accounts or API keys needed. It solves the headache of juggling separate test environments by giving you one shared, consistent world. You can even watch scheduled events arrive in real time, like emails or payments, and reset everything with a single click. It is a brilliant way to test integrations without breaking production. Perfect for demos or just keeping your dev loop tight. Grab it and try it out.

🆔 @hackernewsgithubprojects

📦 google/artemis

Artemis: AI Android Automation

Artemis is the mobile automation tool that lets AI coding assistants drive real Android phones like a human. It solves the brittle nature of traditional app testing by using natural language instructions to navigate interfaces, completely eliminating the need to maintain fragile element selectors. The standout feature is its native integration with IDEs like Claude Code, allowing you to describe a bug in plain English and have the agent reproduce it on a physical device, capture crash logs, and return screenshots. By achieving a success rate of over ninety-nine percent on the AndroidWorld benchmark, it proves that reliable, end-to-end phone control is finally within reach for any developer.

🆔 @hackernewsgithubprojects

📦 carnot-tech/consulting-pptx-skill

consulting pptx skill

consulting pptx skill is the presentation tool that removes the robotic feel from AI generated slides. It solves the problem of generic output by forcing the model to follow a strict rulebook built from real consulting feedback. Instead of picking random templates, it constructs each slide using specific HTML components and runs automated checks to catch design errors. You get boardroom quality decks that actually look human made. It turns a simple prompt into a professional pitch deck without the usual fluff.

🆔 @hackernewsgithubprojects

📦 friedrich-m/unimate

UniMate: One Model, Infinite Skeletons

UniMate is the animation model that finally moves any skeleton you can dream of without needing a new model for each one. It solves the problem where animating a dog, a robot, or a snake usually requires separate training and tools. This single system learns from thousands of diverse movements and uses a text prompt to generate realistic motion for bipedal animals, birds, sea creatures, or even rigid objects in real time. You do not need to retrain the model or run slow optimizations for every new character. It simply takes the structure of your asset and the text description, then produces the animation instantly.

🆔 @hackernewsgithubprojects

📦 trieuntu/vietaidetector

VietAIDetector: How to Spot AI-Generated Vietnamese Text

Spot whether a Vietnamese document was written by a human or generated by artificial intelligence using a tool called VietAIDetector. This open-source software solves a growing problem where it is increasingly hard to tell real human writing from machine-generated content in the Vietnamese language. Instead of relying on guesswork, it uses a clever comparison between two specialized language models to find subtle patterns in the text that indicate its origin. The most interesting part is that it can even read scanned PDF documents, automatically extracting the text and analyzing it without any manual typing.

🆔 @hackernewsgithubprojects

📦 gair-nlp/vidaforge

VidaForge: The Video Data Factory

Transform chaotic web videos into clean, training-ready assets using VidaForge. This open research pipeline handles the messy reality of raw footage, filtering out low-quality clips and standardizing formats so your AI models can actually learn from them. It is built for serious work, integrating directly with heavyweights like V-JEPA2 to turn scattered internet media into structured datasets. The real magic is the automation: it saves researchers from spending months on manual cleanup, letting you focus on model architecture instead of video editing. If you are training vision systems, this is the backbone you have been waiting for to get your data pipeline right.

🆔 @hackernewsgithubprojects

📦 masihsultani/whiteboard-animator

Turn Any Sketch Into a Video With This Free Tool

Feed a plain whiteboard sketch into Whiteboard Animator and watch it draw itself stroke by stroke, exactly like a person using a marker. This tool figures out the order a human would use, writing words first, then tracing outlines and filling shapes in natural paths. It is the free engine behind Kinoslide, meaning you can create professional explainer videos without paying for software or needing a powerful graphics card. The most surprising part is that it runs entirely on your computer using just a standard processor, so you can animate your own diagrams privately and instantly.

🆔 @hackernewsgithubprojects

📦 soullesskid/edgemem

How EdgeMem Remembers Everything

EdgeMem is the local memory engine that lets your AI agents recall past conversations without sending data to the cloud. Instead of relying on expensive cloud APIs for every single interaction, this system builds a structured index right on your device using simple local rules and time-based signals. It handles the heavy lifting of finding relevant details from long chat histories entirely offline, only calling a large language model if you explicitly want to generate a final answer. This approach keeps your private conversations strictly private while slashing latency and costs. It is a clever, lightweight way to give software genuine long-term memory that actually works on your own hardware.

🆔 @hackernewsgithubprojects

📦 kazelad/prediction-market-trade-sdk

Prediction Market Trade SDK: One API for All Markets

One library gives you the exact same code commands for trading on Kalshi, Polymarket, Limitless, and fifteen other prediction market platforms at once. The Prediction Market Trade SDK solves the annoying problem where every betting site uses its own weird data format and connection rules. Instead of writing custom code for each one, you just point your Python or TypeScript script at the SDK, pass in your account details, and start buying or selling shares on real-world outcomes. It handles the messy on-chain signing and settlement in the background so you can just check prices and place orders.

🆔 @hackernewsgithubprojects

📦 nahrek/polyledger

PolyLedger: Your Polymarket Data Vault

PolyLedger is the data tool that turns messy prediction market chaos into one clean file. It sucks every market and on-chain trade from Polygon, then drops it into a single local database you can query with SQL. No servers, no complex setup, just one file on your disk. The coolest part is that if your internet cuts out, it remembers exactly where it left off and picks right up where it stopped. It is basically a time machine for trading data. Now you can actually see who bought what and when, all in one place. Grab it, run it, and stop guessing.

🆔 @hackernewsgithubprojects

📦 livetennisapi/polymarket-tennis

Polymarket Tennis: Live Market Data

Polymarket tennis is the Python toolkit that finally connects prediction market prices to live tennis scores in real time. It solves the headache of manually matching betting markets to actual matches by automatically pairing Polymarket odds with live score feeds, even handling tricky name variations. The coolest part is the joined view: you see the market price sitting right next to the live score, server status, and break point flag, all updated in one snapshot. It is strictly observe-only, meaning it never places trades, so it is a perfect foundation for building your own paper trading bot.

🆔 @hackernewsgithubprojects

📦 cleverfuxaqo1668/polymarket-telegram-bot

Polymarket Telegram Bot: One-Tap Whale Alerts

This Telegram bot spots massive trades on Polymarket and lets you copy them instantly, skipping the slow confirmation popups that often freeze or glitch in the official website. It solves a real annoyance: by the time you manually sign a bet, the price has usually moved. Here, a live feed catches big trades within seconds, and you tap once to sign directly from the chat. This removes the lag, so you can follow the market in real time without jumping between apps. It turns a stressful, click-heavy process into a single, smooth action.

🆔 @hackernewsgithubprojects

📦 eraserlabs/eraser-diagrams

Eraser Diagrams: The AI Diagram Generator

Eraser Diagrams turns plain JSON into polished, boardroom-ready charts, specifically designed to let AI agents build diagrams without you writing a single line of drawing code. The magic lies in its built-in line routing, which automatically draws smart, obstacle-avoiding arrows between your boxes, so your architecture maps actually look professional instead of messy. By keeping the data separate from the visuals, it lets you tweak styles later without breaking the structure. It is a clean, open-source tool for developers who want their AI to generate clear, consistent visuals that are easy to edit and share.

📰 https://news.ycombinator.com/item?id=49363869

🆔 @hackernewsgithubprojects

📦 clipboardhealth/groundcrew

Groundcrew: Your Local AI Task Dispatcher

Groundcrew turns your task list into a fleet of local coding agents that actually work. It grabs tickets from Linear, spins up an isolated environment for each one, and launches a real AI agent in a terminal you can watch live. The best part is the isolation. Every task gets its own branch and sandbox, so you can run five jobs at once without them crashing into each other. It is not a black box. You see the agent code, you can nudge it, and it leaves you a clean pull request when it finishes.

🆔 @hackernewsgithubprojects

📦 jameslhw/polymarket-quick-trade

Polymarket Quick Trade: The Smartest Way to Copy Crypto Whales

Polymarket Quick Trade is the trading tool that finally shows you exactly where the big money is sitting. It scans the blockchain directly to track real-time profit rankings of top traders, bypassing the usual data limits that hide their actual positions. This lets you see who is winning on the latest Bitcoin and Ethereum markets instantly. The interface keeps things simple, using keyboard shortcuts so you can execute trades in seconds without clicking through menus. It is a powerful way to follow smart money without the guesswork. If you want a clearer view of the market, this is the tool to try.

🆔 @hackernewsgithubprojects

📦 tokenrhythm/neohorse

How NeoHorse Learns to Improve Itself

NeoHorse is the language model project that treats AI training as a self-improving loop. Instead of just building a smarter chatbot, the team created a system where the models actually do the work. They run tasks, record what goes right or wrong, and then use those results to update their own training data. It is like a student who grades their own homework and uses the corrections to study better. This creates a cycle where the system gets sharper at every step without needing a human to manually rewrite every lesson. The goal is not just a static tool, but a prototype that can keep getting better on its own.

🆔 @hackernewsgithubprojects

📦 yinminghuang/awesome-agentic-visual-generation-model

Awesome Agentic Visual Generation Model

Sort the chaos of visual AI by asking where a system truly takes the wheel. The awesome agentic visual generation model provides a clear ladder from fixed recipes to systems that learn from past work. It strips away hype to judge if a model actually changes its plan after seeing the result. This helps researchers spot genuine progress in creative tools. You finally get a fair way to measure real agency in visual generation.

🆔 @hackernewsgithubprojects

📦 nehanth/swarmllm

SwarmLLM: Run a 27B Model in Your Browser Tabs

A twenty-seven billion parameter AI model, one that normally requires a powerful server to run, is currently generating text by splitting its workload across a MacBook and an iPhone sitting in the same room. SwarmLLM is a project that takes a massive language model and slices it up so that your laptop, phone, and tablet each hold a small piece of the brain. No installation required, no cloud server, and no data leaves your house. It uses direct peer-to-peer connections between your browser tabs to pass tiny slivers of information back and forth, effectively turning your living room devices into a single, cooperative supercomputer.

📰 https://news.ycombinator.com/item?id=49602451

🆔 @hackernewsgithubprojects

📦 cpldcpu/pico-faces

A Face Generator on a One-Dollar Chip

Pico Faces is the one-dollar microcontroller project that generates human faces using a generative AI model. This tiny chip runs a latent flow diffusion transformer to create one hundred twenty-eight by one hundred twenty-eight pixel images of people in as little as five seconds. It solves the problem of running heavy AI on hardware with almost no memory by squeezing a complex model into just a few million parameters. The real surprise is that it can display these faces on a simple monitor or stream them over USB directly from the chip. You get a fully working AI image generator on hardware that costs less than a coffee.

📰 https://news.ycombinator.com/item?id=49489464

🆔 @hackernewsgithubprojects