Weight quantization and KV cache quantization target two different memory pools. A practical tour of the 2025–2026 state: FP8, FP4, AWQ, GPTQ, GGUF, KIVI, and how to combine them.

Why I moved my web portfolio off Lovable and how I replaced it

Complete guide to setting up vLLM on the AMD Ryzen AI Max+ 395 with unified memory, using the Strix Halo toolbox and TheRock ROCm nightly builds

Complete guide to installing and tuning SGLang for 2x NVIDIA RTX PRO 6000 Blackwell GPUs with MiniMax M2 NVFP4 quantization

Complete guide to setting up NVIDIA GPU passthrough in Proxmox VE 9.1 with a Fedora Server 43 VM for local AI inference

Set up AIBrix on Fedora 42 with NVIDIA GPUs using Minikube None driver for local LLM inference at scale

How to install and configure .NET SDK 10.0, PAC CLI, and the secret service backend needed for Power Apps Code Apps on WSL.

A step-by-step guide to set up vLLM for running local LLMs on Linux with Blackwell architecture support

How to host your vibe-coded lovable site, for free

Short but sweet introduction to "screen"

Learn how to set up automatic deployments from GitHub repositories to Cloudflare Pages for fast, secure web hosting.

A comprehensive guide to setting up and upgrading a Raspberry Pi 4, including network discovery for SSH access and booting from USB SSD for better performance.

Short but sweet instructions on how to install Claude Code and setup integration with GLM 4.6

Basic instructions for developing Power Apps Code Apps in WSL (superseded by newer post)

Work in progress on my Homelab