A day in AI

Projects

Open-source AI that is taking off this week.

Rising on GitHub

vercel-labs/ui

Vercel released UI SDK, an open-source library that lets AI agents build user interfaces with keyboard navigation, focus, selection and drag-and-drop behaviors, with adapters for React, Vue, Svelte, Solid and Web Components.

★ 4,744 +4,642 this weekTypeScript toolsopen source

ZhuLinsen/daily_stock_analysis

Open-source system using LLMs to analyze stock markets across multiple sources with real-time news, automated alerts, and a web dashboard. Runs free on GitHub Actions or locally, with optional paid data sources for stability.

★ 72,313 +3,618 this weekPython open sourcetools

pingdotgg/t3code

T3 Code is an open-source control plane that lets developers manage multiple AI coding agents (Claude Code, Codex, Cursor, Grok Build and others) from one interface across desktop, web and phone, with built-in Git integration and remote access.

★ 61,034 +2,909 this weekTypeScript open sourcetools

expo/exact

Expo released Exact, an open-source framework that lets developers write a single app that runs natively on web, macOS, iOS, and Linux, with support for AI agents to build, run, and test it across all platforms.

★ 4,236 +1,606 this weekRust open sourcetools

ChromeDevTools/chrome-devtools-mcp

Chrome DevTools MCP is an open-source Model-Context-Protocol server that lets coding agents like Claude and Cursor control and inspect a live Chrome browser for debugging, performance analysis, and automation.

★ 56,617 +1,599 this weekTypeScript toolsopen source

QwenLM/qwen-code

Qwen Code is an open-source AI coding agent that runs in terminals and editors, supporting multiple LLM providers and featuring agentic capabilities like auto-memory and agent teams without setup required.

★ 30,506 +1,375 this weekTypeScript toolsopen source

RunanywhereAI/trnrt

TrnRT is an open-source inference engine for large language models on AWS Trainium chips, offering an OpenAI-compatible API, 1M-token context support, and costs $0.167 per million tokens on capacity blocks.

★ 1,586 +1,329 this weekPython open sourceinfrastructure

coreyhaines31/marketingskills

Marketingskills is an open-source collection of AI agent skills for marketing tasks—conversion optimization, copywriting, SEO, analytics, and growth engineering—built for Claude Code and other coding agents. Works with major AI coding tools and integrates with marketing platforms.

★ 60,459 +1,131 this weekJavaScript open sourcetools

elder-plinius/CL4R1T4S

An open-source repository collects extracted system prompts from major AI models including ChatGPT, Claude, Gemini, and others to show how AI systems are instructed to behave.

★ 53,066 +1,063 this week open sourcesafety

obsidianmd/obsidian-skills

Open-source Obsidian skills enable AI agents to interact with Obsidian notes and files. Developers using Claude Code, Codex, Open Code or similar agent frameworks can install these skills to extend their agents' capabilities.

★ 52,725 +1,045 this weekShell open sourcetools

ruvnet/RuView

RuView uses WiFi signals to detect presence, breathing, heart rate, and activity through walls without cameras. It integrates with smart-home platforms and runs locally on ESP32 boards costing under $10.

★ 104,337 +1,015 this weekRust open sourcetools

kajisho5/ffmpeg-skill

kajisho5/ffmpeg-skill is an open-source set of 42 FFmpeg tools with machine-readable contracts and an MCP server for AI coding assistants like Claude Code and Cursor. It enables local video editing—cutting, adding captions, reframing, adjusting loudness and exporting—without cloud services or API keys.

★ 5,795 +916 this weekPython toolsopen source

FrigadeHQ/yogurt

Yogurt is an open-source desktop app that records meetings, transcribes speech and generates summaries using local AI models, keeping all data on your computer.

★ 1,563 +828 this weekTypeScript open sourcetools

wici-ai/inferssd

InferSSD, an open-source tool built on llama.cpp, lets developers run large mixture-of-experts models from SSD storage that exceed available RAM or VRAM, supporting models like DeepSeek and Qwen via GGUF files with a browser chat and OpenAI-compatible API.

★ 1,160 +789 this weekC++ open sourcetools

f/prompts.chat

f/prompts.chat is an open-source library for sharing and discovering AI prompts that work with ChatGPT, Claude, Gemini and other models. Users can self-host it for private use, browse community bots and bot teams, and learn prompt engineering through an interactive guide.

★ 177,487 +767 this weekHTML open sourcetools

agentes-chat/agentes-chat

Open-source self-hosted Slack alternative lets teams collaborate with AI agents powered by Claude, ChatGPT, Gemini or Grok, with browser, computer and tool access, plus web, desktop and mobile apps.

★ 1,459 +643 this weekTypeScript open sourcetools

PurpleDoubleD/locally-uncensored

PurpleDoubleD/locally-uncensored is an open-source desktop application that runs chat, image generation, video generation and a coding agent locally without cloud services, Docker or terminal access. Available for Windows and Linux.

★ 4,264 +629 this weekTypeScript open sourcetools

Gentleman-Programming/gentle-ai

Gentle-AI is an open-source tool that configures AI coding agents like Claude Code and Cursor with features including persistent memory, skills, MCP servers, and personas, without locking users into a specific agent.

★ 9,205 +550 this weekGo open sourcetools

synchopate/llama-modchip

llama-modchip is an open-source repository that enables running 284B DeepSeek-V4-Flash and 30B mixture-of-experts models on 64GB of CPU RAM without GPUs, using optimizations like SwiGLU-H2 and Stockfish-PVS.

★ 1,258 +505 this weekPython open sourcetools

agent-of-empires/agent-of-empires

Open-source repository for managing multiple Claude Code and OpenCode agents through terminal or web UI, with support for other coding agents including Mistral, Gemini, and Copilot CLI.

★ 4,461 +503 this weekRust open sourcetools

Open models on Hugging Face

MiniMaxAI/MiniMax-H3-Image

MiniMaxAI released MiniMax-H3-Image, an open-source text-to-image and image-editing model that generates 16-megapixel images in 6 steps with support for up to 16 reference images. Developers and creators can use it for high-resolution image generation and editing.

text to image♥ 270 ↓ 667 open sourcemodels 12 days ago

chatrlabs/Qwen-Image-2.1-Uncensored-Omni-Quantized

ChatR Labs released Qwen-Image-2.1-Uncensored-Omni-Quantized, a quantized text-to-image model on Hugging Face with multiple weight formats for different inference frameworks. Developers can use it for text-to-image generation and image editing with ComfyUI or Diffusers.

text to image♥ 291 ↓ 96,251 open sourcetools 12 days ago

higurashi-ai/Haru-4B

Haru-4B is a 4 billion parameter multilingual speech-to-speech model that listens and speaks simultaneously like a phone call, released open-source for research. Developers can customize its character and voice.

audio to audio♥ 232 ↓ 1,262 open sourcemodels 3 days ago

Edge0/Edge0-145B-NVFP4

Edge0-145B is a 145-billion-parameter open model for text generation using mixture-of-experts and 4-bit quantization techniques. It targets developers and researchers working with large language models on resource-constrained inference.

text generation♥ 241 ↓ 33,819 open sourcemodels 5 days ago

adventists-ai/Sophon-Realtime-v1

Sophon-Realtime v1 is an open full-duplex voice model that conducts natural conversations by listening continuously, detecting paralinguistics like tone and emotion, and responding with matching expressions. Developers can self-host or use a hosted API compatible with OpenAI Realtime.

any to any♥ 270 open sourcemodels 9 days ago

ConwayResearch/Underdog-Woof

Underdog-Woof is a 40.6 MB English text-to-speech model with two voices that runs on CPU with no GPU or API key, achieving 1.42% word error rate and 13.8x faster-than-real-time speed on standard hardware.

text to speech♥ 256 ↓ 10 modelsopen source 10 days ago

thinkingmachines/Inkling-2-Small

Inkling 2 Small is an open-weight multimodal model accepting text and image inputs to generate text, designed for developers building conversational AI, coding assistants, chatbots and retrieval-augmented generation systems.

image text to text♥ 211 ↓ 1,064 open sourcemodels 11 days ago

Cactus-Compute/mirage

Cactus-Compute released mirage, an open vision model on Hugging Face for image understanding and structured extraction tasks including object detection and segmentation.

image text to text♥ 214 ↓ 16,133 open source 10 days ago

unsloth/Qwen3.8-27B-GGUF

Qwen3.8-27B is a 27-billion-parameter open model with vision capabilities, designed for coding, research, and multi-step tasks. The GGUF quantization from Unsloth claims 5-10% better accuracy than other quantizations.

♥ 5,795 ↓ 6,097,719 modelsopen source 13 Aug 2026

Zyphra/ZONOS3-Mini

Zyphra released ZONOS3-Mini, an open-source text-to-speech model with 100 million parameters that runs on phones, laptops and browsers. It supports voice cloning, emotional control and multilingual output.

text to speech♥ 222 ↓ 8,543 open sourcemodels 7 days ago

Viggle/Viggle-Remake

Viggle-Remake is a video-to-video model for editing films: recasting characters, swapping outfits, adding or removing objects, and changing weather or style. It works with coding agents or manual prompts.

video to video♥ 194 ↓ 76,308 open sourcetools 24 days ago

Comfy-Org/MiniMax-H3-Image

Comfy-Org/MiniMax-H3-Image is a text-to-image model on Hugging Face. Commercial use requires MiniMax authorization if annual revenue exceeds $1 million or the service is hosted for third parties.

text to image♥ 188 ↓ 18 open source 7 days ago