Discover the Best AI Resources
Curated essentials, no noise — just what matters
A line-by-line PyTorch reimplementation of the Transformer paper as a runnable notebook, where each part of the paper sits next to the code that implements it — turning a dense architecture into something you can read and run end to end.
AI Infra2022
Provides runtime support for reversible effects and reactive coeffects so components can be declared, composed, and hot-replaced safely; includes effect tracking, coeffect resolution, a declarative component loader and HMR — aimed at plugin-driven agent harnesses and dynamic systems.
AI Infra2022
Fused CUDA kernels that compute exact attention without ever writing the full N×N score matrix to GPU memory, cutting memory from quadratic to linear and speeding up training and inference on A100/H100. Ships FlashAttention-2/3 plus KV-cache decode paths.
AI Video2022
Generate short social videos from Reddit threads in one command — captures post content, assembles visuals and optional TTS narration, and outputs an upload-ready MP4. Runs locally with Python + Playwright; does not auto-upload for safety.
Builds a single rigorous theory from one question: why some bit strings look random. Defines plain and prefix complexity, the incompressibility method, and Martin-Löf randomness, tying information content to whether a short program can reproduce a string.
AI Model2022
Transformer-based foundation model for tabular data that provides pre-trained checkpoints for fast classification and regression, with GPU-accelerated local inference and an optional cloud client. Best suited for small-to-medium datasets (~≤50k rows).
AI Model2022
Runs pretrained diffusion models for image, video, and audio generation through composable pipelines. It separates pipelines, schedulers, models, adapters, and memory optimizations so teams can prototype quickly without locking into one model family.
AI Infra2022
Collects metrics, distributed traces, and continuous profiles via eBPF with zero code instrumentation, covering apps in any language plus gateways, service meshes, databases, and queues. Profiling adds under 1% overhead.
AI Image2022
Enlarges and enhances low-resolution images using AI models (Real-ESRGAN) through a cross-platform desktop app. Runs on a local NCNN/Vulkan backend (requires a Vulkan-compatible GPU), offers an Electron GUI plus a CLI backend (upscayl-ncnn), and supports custom models for different image types.
Chatbot2022
Runs a local AI assistant across WeChat/Feishu/DingTalk/WeCom/QQ/MP/Web, with an Agent mode for task planning, long-term memory, Skills, and tool calling so it can keep working toward goals rather than just chat.
AI Agent2022
Framework for building multi-channel AI assistants that autonomously plan tasks, invoke tools/skills, and keep long-term memory; supports many LLM providers and channels (WeChat, Feishu, QQ, web) for local or server 24/7 deployment.