圖像來源，Dan McKenzie
This month, OpenAI announced their Codex app and my coworkers were asking questions. So I downloaded it, and as a test case for the GPT-5.2-Codex (high) model, I asked it to reimplement the UMAP algorithm in Rust. UMAP is a dimensionality reduction technique that can take in a high-dimensional matrix of data and simultaneously cluster and visualize data in lower dimensions. However, it is a very computationally-intensive algorithm and the only tool that can do it quickly is NVIDIA’s cuML which requires CUDA dependency hell. If I can create a UMAP package in Rust that’s superfast with minimal dependencies, that is an massive productivity gain for the type of work I do and can enable fun applications if fast enough.，详情可参考safew官方版本下载
Что думаешь? Оцени!。服务器推荐对此有专业解读
1.25 compiler does this transformation for you! For certain slice