Blog

Thoughts on infrastructure, AI, homelab, and tech.

9 May 2026

We Built a World Cup Trip Command Centre Because Chaos Was Getting Ideas

A pocket mission board for football, flights, hotels, costs, documents, and the ancient group-trip ritual of someone asking a question answered three messages ago.

We Gave Hermes a Dreaming Mode Because Apparently Stealing Good Ideas Is Also Intelligence

A lab note on borrowing OpenClaw's dreaming mode idea, wiring it into Hermes properly, and then discovering that every ambitious memory system is still one bad cron edge case away from humiliation.

Teaching the Lab a New Trick: Bringing the Strix Halo NPU Online

We’ve been building a dedicated NPU lane in the lab on Strix Halo — not as a gimmick, but as a serious path for efficient always-on inference, lighter agent jobs, and future local AI workflows. Here’s what’s already working, what took actual effort, and what’s still blocking first real NPU inference.

Turning Conversations Into Action: What We’ve Been Building Around Omi

How we’re turning messy transcripts into cleaner actions, sharper briefs, and follow-ups that are actually useful instead of mechanically tidy and semantically hollow.

Adding a 3090 to the Lab Without Summoning Infrastructure Goblins

A used RTX 3090 is still one of the most interesting homelab GPU upgrades you can make. Here’s why I added one, why I paired it with a Strix Halo box, and how to think about splitting workloads cleanly between CUDA and AMD graphics without turning your setup into a confused science project.

Giving Bill and Homer Superpowers: The Skill Design That Fixed Our Coding Loop

Bill writes code. Homer reviews it. Both were making avoidable mistakes — not because the models were bad, but because they had no structured workflow. We fixed that with two tiny skill files. Here's the logic.

The Forge: A PR Factory Powered by Local LLMs

I built a system where local AI agents write code, open PRs, review them, and merge them — autonomously, overnight, while I sleep. Here's how The Forge works.

Building a Self-Deploying Agent Dashboard

The agents built their own dashboard today. Bill has shipped 13 PRs so far — 5 today alone. Homer reviews them, Bob deploys. Toast notifications, system health monitoring, and cron status cards — all shipped by local LLMs on a Strix Halo.

Coding Loop v2: From Chaos to Simplicity

Started the day with fragile multi-board automation. Ended with a clean 3-actor loop, a pixel art dashboard, and 6 tasks queued for overnight cooking.

Lemonade + Qwen3.5: The Easiest Way to Run a 35B MoE on AMD Hardware

Forget building llama.cpp from source. Lemonade gives you a dead-simple OpenAI-compatible inference server for AMD GPUs — and Qwen3.5-35B-A3B is the perfect model to run on it.

Building a Local AI Agent Fleet on AMD Hardware

A full day optimising a multi-agent AI fleet — fixing tool calling, building a live dashboard, investigating AMD Lemonade, and poking at the Strix Halo NPU.

Running Ollama + ROCm on AMD Strix Halo in a Proxmox LXC

AMD's Strix Halo silicon (gfx1151) is fast, weird, and very new — too new for most pre-built Ollama binaries. Here's the full process to get ROCm 7.2 working in a Proxmox LXC, including every gotcha I hit along the way.