📦 kuleshov-group/setdlms

Set Diffusion

AI language models usually generate text strictly from left to right, but a clever new approach lets them write in flexible, parallel chunks to speed up generation. This repository contains the official code for setdlms, a project that bridges the gap between traditional step-by-step text generation and fully parallel diffusion. By dividing text into variable-sized groups of words, it lets developers tune the balance between generation speed and output quality. It is a fantastic tool for anyone experimenting with faster text summarization or mathematical reasoning models. Check out the project to start blending autoregressive and diffusion techniques on your own hardware.

🆔 @hackernewsgithubprojects

📦 yale-nlp/rlmf

RLMF

Teach language models to stop bluffing when they do not know the answer. While AI is great at generating text, it is notoriously bad at knowing when it is actually wrong. That is where rlmf comes in. By rewarding models during reinforcement learning based on how accurately they self-assess their own performance, this framework teaches them genuine self-awareness. Instead of confidently hallucinating, the AI learns to scale back its confidence when it is guessing. It is a brilliant way to make AI trustworthy, giving you a partner that finally knows its own limits.

🆔 @hackernewsgithubprojects

📦 coderluii/holyclaude

HolyClaude

HolyClaude is the self-hosted AI development workstation that packages everything you need to build with Claude Code into a single, ready-to-run container. If you have ever tried to manually configure a headless browser, set up multiple terminal AI clients, and get Playwright and Chromium working inside Docker without permission errors, you know it is a massive chore. This project solves that headache by bundling a full web interface, eight different AI command-line assistants, and over fifty pre-installed developer tools into one simple package. The absolute best part is that it lets you run the real, official Claude Code CLI in your web browser using your existing subscription or API key...

🆔 @hackernewsgithubprojects

📦 shridarpatil/whatomate

Whatomate

Whatomate is the open-source WhatsApp Business platform that runs entirely as a single binary app, giving you complete control over your customer communication without complex hosting. Built with Go and Vue, it lets you connect directly to the WhatsApp Cloud API to manage multi-tenant setups, design automated chatbots, and handle real-time team chats. The most impressive part is its full-featured voice calling system, packed with interactive voice menus, call transfers, hold music, and automatic recordings. It is an incredibly powerful, self-hosted communication hub. Grab the Docker image or the pre-built binary to launch your own self-hosted setup today.

🆔 @hackernewsgithubprojects

📦 x011/smtp-tunnel-proxy

SMTP Tunnel Proxy

smtp-tunnel-proxy is the clever tunnel that hides your internet traffic inside fake email conversations to slip past strict firewalls. To a deep packet inspection firewall, your connection looks exactly like a normal Postfix mail server negotiating a secure email delivery over standard ports. Once that realistic handshake is complete, the system instantly switches to a high-speed binary streaming mode that runs at full network bandwidth. By wrapping your data in standard email security encryption, it makes your browsing completely invisible to network blocks. It is a brilliant way to keep your connection secure and open without raising any red flags.

🆔 @hackernewsgithubprojects

📦 ysharma3501/luxtts

LuxTTS

Clone any voice with incredible realism and run it instantly on your own hardware using luxtts. This lightweight tool generates crystal-clear forty-eight kilohertz speech at over one hundred and fifty times realtime speed, making most other voice models look sluggish. It is highly optimized, fitting comfortably within just one gigabyte of video memory, so you can easily run it locally on your own graphics card or even a standard processor. By refining the underlying architecture into just four quick generation steps, it delivers incredible audio quality without demanding a massive server. It is the perfect setup for fast, local, and high-quality voice generation.

🆔 @hackernewsgithubprojects

📦 outsourc-e/hermes-workspace

Hermes Workspace

You can now run a complete visual command center for AI agents directly in your web browser. hermes-workspace provides a native web environment that brings chat interface, terminal control, memory management, and specialized agent skills into one unified dashboard. Instead of trying to manage complex AI workflows through separate, disconnected command lines, this project puts everything into an interactive workspace. You can watch your agents run tasks, inspect their logic, and interact with their shared memory in real time. It is the perfect tool for developers who want to build, debug, and orchestrate cooperative agent networks without losing sight of what is happening under the hood.

🆔 @hackernewsgithubprojects

📦 tradermonty/claude-trading-skills

Claude Trading Skills

This open-source toolbox turns your AI assistant into a rigorous investment partner by teaching it a massive library of modular trading skills. Built specifically for Claude Code, it provides highly structured markdown templates that define precise analytical behaviors, from spotting market tops and evaluating dividend growth pullbacks to analyzing chart setups. Instead of letting an AI make blind guesses, you feed these specialized playbooks into your assistant to run disciplined market routines, screen stocks, manage portfolio risk, and conduct postmortems on your trades. It is the ultimate framework to build a structured, repeatable investing process without coding automated bots from scratch.

🆔 @hackernewsgithubprojects

📦 foru17/neko-master

Neko Master

See your network traffic with absolute clarity using a gorgeous, modern dashboard designed for local gateway environments. It connects to your gateway, captures live traffic, and visualizes everything from domain and proxy performance to geographic connections in real time. The coolest part is how it handles multiple locations: instead of making your central panel connect directly to remote gateways, lightweight agents run locally next to each gateway and safely feed the data back, keeping everything secure and private. It is the ultimate tool for curious minds who want to understand exactly where their home or office data is flowing.

🆔 @hackernewsgithubprojects

📦 llmsresearch/paperbanana

PaperBanana

PaperBanana is the automated illustration tool that finally turns complex research descriptions into publication-ready academic figures and diagrams. If you have ever tried to design a clear methodology diagram or a statistical plot for a paper, you know how incredibly tedious it is. This clever tool uses a multi-agent AI pipeline to read your text description or PDF, plan the visual layout based on style guides from major conferences, and generate your diagram. It then automatically critiques and refines the image over multiple rounds until it looks clean and professional. You can even run a local web interface to manage your figures, generate complex plots directly from data tables, or...

🆔 @hackernewsgithubprojects

📦 kittendevv/invio

Invio

You can now run your entire billing system on your own hardware without dealing with expensive subscriptions or cluttered menus. Invio is a self-hosted, ultra-lightweight invoicing platform that cuts out the bloated setup wizards and corporate fluff entirely. It runs on a SvelteKit frontend and a simple SQLite backend, allowing you to generate professional invoices, get paid, and securely share links with clients who do not even need to make an account. You retain complete control over your data, free from sudden pricing hikes or service shutdowns. It is the perfect, straightforward way to handle billing on your own terms.

🆔 @hackernewsgithubprojects

📦 vast-ai-research/anigen

AniGen

Turn any flat, two-dimensional picture into a fully rigged, animation-ready 3D character with just a single click. Instead of spending hours meticulously shaping a model, building its skeleton, and painting complex skinning weights, this clever tool generates the entire package all at once. By predicting the shape, bone structure, and movement joints together in one unified step, it creates highly consistent assets that actually move naturally. You can feed it photos of animals, humans, or even complex machinery, and grab a standard file ready to animate immediately. It is an absolute game-changer for bringing static designs to life instantly.

🆔 @hackernewsgithubprojects

📦 firecrawl/open-scouts

Open Scouts

Open Scouts is the AI-powered web monitoring platform that finally automates your search routines without any manual hassle. Instead of wasting hours refreshing tabs, you simply create custom scouts to track specific topics like local restaurant openings or AI industry news on a schedule you choose. The system runs these tasks in isolated edge functions, searching and scraping the web behind the scenes, and then translates the results into clear, one-sentence summaries. When it finds something relevant, it automatically alerts you via email. It is the perfect, stress-free way to stay informed on autopilot.

🆔 @hackernewsgithubprojects

📦 agi-now/buffett-skills

Buffett Skills

Supercharge your investment research by analyzing companies like Warren Buffett using AI. This project lets you install a custom suite of Warren Buffett investment thinking skills directly into Claude Code. When you ask about any business, financial report, or stock, it automatically triggers a systematic deep dive based on forty-nine key concepts compiled straight from Buffett's letters. It runs a quick eight-question filter on things like competitive moats, pricing power, and management integrity, and then calculates intrinsic value with a margin of safety. You get objective, structured buy-or-sell breakdowns delivered right in your terminal.

🆔 @hackernewsgithubprojects

📦 jamesob/local-llm

local-llm

local-llm is the hardware and configuration blueprint that shows you how to build a supercomputer in your home to run massive artificial intelligence models privately. Instead of paying ongoing fees to big cloud providers, this project details how to construct a physical rig using smart hardware workarounds, like using specific PCIe switches so multiple graphics cards can talk directly to each other at lightning speeds without needing a multi-thousand-dollar motherboard. It covers everything from custom wooden enclosure designs to power-limiting tricks that let you run high-end hardware on a standard household outlet. It is the ultimate guide to taking control of your own data.

📰 https://news.ycombinator.com/item?id=48775921

🆔 @hackernewsgithubprojects

📦 celadaniel/free-ai-resources-x

Free AI Resources X

Finding high-quality, free machine learning guides often means digging through unorganized link dumps, but free-ai-resources-x changes that. This curated repository acts as an open-source gateway to learning artificial intelligence, featuring over four hundred verified, completely free resources organized by difficulty. Instead of random links, it provides clear, step-by-step learning paths for everyone from absolute beginners studying basic Python to experts diving into multi-agent systems and robotic controls. It is a structured map of high-quality courses, books, and tools from top-tier institutions. If you want to build a real foundation in AI without spending a dime, this repository is the perfect place to start.

🆔 @hackernewsgithubprojects

📦 teamchong/pxpipe

pxpipe

An ingenious local proxy called pxpipe is cutting AI API costs by up to seventy percent by turning bulky text into images before it leaves your machine. While large language models charge a premium to read massive blocks of text like system prompts, tool documentation, and long chat histories, their price for reading an image is fixed by its pixel dimensions. By pack-rendering dense text into compact PNGs, the model reads the visual data at a fraction of the token cost. It automatically evaluates each request, only converting when the math saves you money, and leaves your actual output untouched.

📰 https://news.ycombinator.com/item?id=48776464

🆔 @hackernewsgithubprojects

📦 teamaiq/colab-notebooks

colab-notebooks

colab-notebooks is the curated resource hub that finally makes high-end artificial intelligence tools accessible to anyone directly through a standard web browser. Instead of wrestling with complex local setups, expensive graphics cards, or difficult installations, you get pre-configured, one-click templates optimized specifically for free cloud servers. You can immediately run advanced voice cloning, generate high-fidelity music, or create detailed videos using massive open models without spending a dime. It is the perfect playground for curious creators who want to experiment with next-generation media generation instantly. Grab a notebook and start creating today.

🆔 @hackernewsgithubprojects

📦 symgraph/idassist

IDAssist

Supercharge your reverse engineering workflow by bringing advanced AI analysis directly into IDA Pro. This clever plugin integrates directly into your workspace to automatically translate dense assembly or pseudocode into clear, natural language explanations. Instead of manual guesswork, you can generate instant security risk profiles, get smart rename suggestions for variables, and even build an interactive knowledge graph of the entire binary. It can even run an autonomous reasoning agent to investigate complex code paths for you. It is the ultimate sidekick for making sense of complicated binaries without losing your focus.

🆔 @hackernewsgithubprojects

📦 brobertsaz/claude-os

Claude OS

Give your AI assistant a permanent memory so you never have to repeat yourself again. If you are tired of explaining your codebase architecture, api tokens, and coding standards to Claude every single time you open a new terminal session, this project fixes that by introducing a local knowledge base that automatically tracks your decisions. It uses a lightning-fast structural indexing system to map your entire repository in under thirty seconds without burning through api credits, and lets you save crucial developer insights simply by talking naturally. Now, Claude remembers your choices across different days and projects, turning your AI assistant into a true long-term partner.

🆔 @hackernewsgithubprojects