I write about what I have built around Anthropic's Claude Code as a personal operating environment: the architecture, the incidents that shaped it, what I would do differently, and the specific patterns that have earned their place.
Every post is grounded in something concrete. No speculation about the future of AI.
Writing
Long-form posts. Architecture, incidents, the patterns that earned their place.
- Context as a first-class artifact: the /deep-context pipeline
- "Email me when done": a persistent task runner with a delivery guarantee
- Lessons as code: turning postmortems into pre-flight checks
- Memory that sleeps: a tiered memory architecture with daily consolidation
- One hour, one command: disaster recovery for a solo AI shop
- A personal AI operating environment: worked example and receipts
- Five things I built to help my AI agent that I had to remove
- Six layers of defence for an AI agent over a 3D printer
- Three-way AI model debate as a pre-commit gate: receipts from several months of use
- What it means to be AI-native
Stories
Shorter, first-person pieces about specific builds and what they taught me.
- 918MB, an Ofsted inspection, and a governor who is not a developer
- Building from my phone while watching the kids
- From model to agent: what changed when I stopped predicting and started investigating
- One hour, one marketing list
- The dragon my daughter designed
- The hackathon that started it all