5 Must‑Try Open‑Source Projects: 3D Tetris, Code Analyzer, AI Notebook & More

1. Open‑Source 3D Tetris

This project recreates the classic Tetris game in a three‑dimensional browser environment using React for the UI and Three.js for 3D rendering. It serves as a concrete example for developers interested in combining React with Three.js to build interactive 3D web games.

https://github.com/RylanBot/threejs-tetris-react

2. "fuck‑u‑code" Code‑Quality Analyzer

Named provocatively, this tool scans a given codebase and evaluates it across seven dimensions: cyclomatic complexity, function length, comment coverage, error‑handling completeness, naming conventions, duplication, and overall structural soundness. It outputs a 0‑100 “shit‑mountain index” where higher scores indicate poorer quality.

https://github.com/Done-0/fuck-u-code

3. Open Notebook – Open‑Source Alternative to Google NotebookLM

Open Notebook replicates the functionality of Google’s NotebookLM, offering a multimodal research workspace that can ingest PDFs, videos, audio, webpages, and Office documents. Its interface splits into three panes: a left pane for source material, a central pane for notes, and a right pane for contextual AI chat. It also features a podcast‑generation module that can create multi‑host scripts with configurable styles.

The platform runs locally, giving users full control over data and models, which is crucial for privacy‑sensitive research.

https://github.com/lfnovo/open-notebook

4. Kode – AI Assistant in the Command Line

Kode is an open‑source AI companion that runs directly in the terminal. It can understand codebases, edit files, execute shell commands, and handle complex development tasks, similar to Claude Code. Kode can dynamically switch among multiple large language models, using an internal task‑allocation strategy to select the most suitable model for each request.

The project is authored by Xinlu Lai, a contributor to the Chinese Llama 3 model.

https://github.com/shareAI-lab/Kode

5. Meta’s DINOv3 Visual Model Family

Meta released the official PyTorch reference implementation and pretrained checkpoints for the DINOv3 visual foundation models. These models deliver high‑quality, generic image embeddings that work zero‑shot across tasks such as classification, segmentation, and depth estimation. Pretrained weights are available for ViT and ConvNeXt architectures and can be loaded via PyTorch Hub or Hugging Face Transformers.

https://github.com/facebookresearch/dinov3