Articles
Long-form professional and technical writing.
Homelab AI: Squeezing a 64k Context Window into 8GB VRAM (at 50 TK/s)
How to bypass VRAM limits and run Qwen 3.5 9B with a massive 64k context window on an old 8GB RTX 2080. No model quantization required—just pure Docker environment and KV cache optimization delivering 50 tokens per second inside VS Code via Continue.
Is It Easier to Be Language Agnostic in 2026?
The short answer is yes. But the reason why says something about what "knowing a language" was ever really made of, and what is left once an AI agent handles the syntax for you.
Homelab: What's New
Catching up on what has changed in the homelab stack since the original write-up: local LLMs via Ollama and Open WebUI, Portainer, and some services that are still on the way.
Building My Homelab Docker Stack
A walkthrough of the self-hosted Docker stack running on my NAS: media, DNS, authentication, CI/CD, monitoring, and a NVIDIA GPU doing actual work.
Introducing AI to Development Teams
After a short but very productive evaluation period, I introduced AI tooling to several development teams at work. Here is what I learned about making that transition stick.
Building an Elite Dangerous Community Platform
SHED is a full-stack .NET platform for our Elite Dangerous group, with a Blazor web app, WPF desktop client, and ASP.NET Core API backend.
Building Jellyfin Plugins with AI
I built two Jellyfin plugins in a week using AI-assisted development. Here is what the process looked like and what I learned along the way.
Building a Type-Safe Twitch API Library in .NET
How I built a modular, type-safe wrapper for the Twitch Helix API using Refit and .NET 10, with full OAuth and EventSub support.
The Recruitment Process
Ten weeks, many interviews, and a lot learned about the current state of tech recruitment in the UK.