Learning diary › Year 2025 › July, 2025 › 2025-07-15 [2025-07-15]

✍️source

- #system
    - A distributed systems reliability glossary
- #gpu
    - Rust running on every GPU (on HN) (on lobste.rs)
        - https://mlir.llvm.org/docs/Dialects/NVGPU/
        - https://mlir.llvm.org/docs/Dialects/AMDGPU/
        - https://mlir.llvm.org/docs/Dialects/XeGPU/
    - Apple's MLX adding CUDA support (on HN)
        - Programs built against MLX -> Can take advantage of CUDA-enabled chips
        - not the other way around
        - SCALE documentation
            - SCALE is a GPGPU programming toolkit that can natively compile CUDA applications for AMD GPUs.
            - spectral-compute/scale-examples
    - GPUHammer: Rowhammer Attacks on GPU Memories are Practical #sec
    - A GPU Calculator That Helps Calculate What GPU to Use (on HN)
- #db
    - Embedding User-Defined Indexes in Apache Parquet Files
- #selfhost
    - Packages proxies
        - EpicWink/proxpi: PyPI caching proxy
        - panamax-rs/panamax: Mirror rustup and crates.io repositories, for offline Rust and cargo usage.
        - Pulp Project

            > Some specialized proxies handle different kinds of packages, like Pulp.

        - soulteary/apt-proxy: [Ubuntu / Debian / CentOS / Alpine] Lightweight APT / YUM / APK CHACHE PROXY just over 2MB in size!
        - NGINX Content Caching | NGINX Documentation

            > Squid is boring; lets use Nginx cache.

- #agent
    - Show HN: We made our own inference engine for Apple Silicon
        - A high-performance inference engine for AI models, in Rust
        - trymirai/lalamo: JAX infrastructure for model optimisation
        - Anemll/Anemll: Artificial Neural Engine Machine Learning Library
- #nix
    - Trying Guix: A Nixer's Impressions
- #workflow
    - Nextflow: System for creating scalable, portable, reproducible workflows
    - Metaflow: Build, Manage and Deploy AI/ML Systems (on HN)
        - Why Metaflow? (on HN)